node-autotask 0.84.0 → 0.85.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (38) hide show
  1. package/CHANGELOG.md +7 -0
  2. package/MCP_TOOL_CATALOG.json +4 -4
  3. package/capabilities.json +179 -162
  4. package/dist/capabilities-data.js +178 -161
  5. package/dist/capabilities.cjs +4 -4
  6. package/dist/capabilities.js +1 -1
  7. package/dist/{chunk-4NLWSHCB.cjs → chunk-7GNZT7Y7.cjs} +22 -11
  8. package/dist/{chunk-Z56J7TFC.js → chunk-EPJG7M5C.js} +42 -7
  9. package/dist/{chunk-HASLVFC7.js → chunk-GTTAPAGP.js} +3 -3
  10. package/dist/{chunk-RWBOJ5WK.js → chunk-HF5YG4TV.js} +2 -2
  11. package/dist/{chunk-MVGGHMOB.cjs → chunk-JCKYM7FD.cjs} +3 -3
  12. package/dist/{chunk-36RXJUSA.cjs → chunk-JV6MX2U3.cjs} +237 -202
  13. package/dist/{chunk-AWK7FIZD.js → chunk-L6YCH2I4.js} +17 -6
  14. package/dist/{chunk-SNILYCAV.cjs → chunk-PBOH2OHJ.cjs} +30 -30
  15. package/dist/{index-BZItfp_H.d.cts → index-CCW_EpLg.d.cts} +140 -140
  16. package/dist/{index-BZItfp_H.d.ts → index-CCW_EpLg.d.ts} +140 -140
  17. package/dist/{index-DdvYwR2w.d.cts → index-DkyNXWyt.d.cts} +17 -4
  18. package/dist/{index-f1c_EdAt.d.ts → index-rWcPjRpX.d.ts} +17 -4
  19. package/dist/index.cjs +84 -84
  20. package/dist/index.d.cts +3 -3
  21. package/dist/index.d.ts +3 -3
  22. package/dist/index.js +5 -5
  23. package/dist/operations/index.cjs +51 -51
  24. package/dist/operations/index.d.cts +4 -4
  25. package/dist/operations/index.d.ts +4 -4
  26. package/dist/operations/index.js +2 -2
  27. package/dist/pool.cjs +12 -12
  28. package/dist/pool.d.cts +1 -1
  29. package/dist/pool.d.ts +1 -1
  30. package/dist/pool.js +4 -4
  31. package/dist/resources/index.cjs +27 -27
  32. package/dist/resources/index.d.cts +2 -2
  33. package/dist/resources/index.d.ts +2 -2
  34. package/dist/resources/index.js +3 -3
  35. package/dist/types/index.d.cts +1 -1
  36. package/dist/types/index.d.ts +1 -1
  37. package/docs/VENDOR-QUIRKS.md +3 -1
  38. package/package.json +1 -1
@@ -427,8 +427,8 @@ const CAPABILITY_DATA = {
427
427
  ],
428
428
  "groups": {
429
429
  "B": {
430
- "planHash": "5a5cfa013a7d398302f69e1c12e260ea29a89942aec743997be1f8cdf60a8df1",
431
- "builtAt": "2026-11-19T23:12:11.551Z",
430
+ "planHash": "9521525945b1306da363ce5fb38649e17c5526f2c0842da64c3bec7993e62735",
431
+ "builtAt": "2026-11-20T05:02:13.666Z",
432
432
  "records": [
433
433
  {
434
434
  "id": "appointments.count",
@@ -821,7 +821,7 @@ const CAPABILITY_DATA = {
821
821
  "params": {
822
822
  "type": "object",
823
823
  "required": false,
824
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
824
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
825
825
  "properties": [
826
826
  {
827
827
  "name": "filter",
@@ -1380,7 +1380,7 @@ const CAPABILITY_DATA = {
1380
1380
  "params": {
1381
1381
  "type": "object",
1382
1382
  "required": false,
1383
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
1383
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
1384
1384
  "properties": [
1385
1385
  {
1386
1386
  "name": "filter",
@@ -1809,7 +1809,7 @@ const CAPABILITY_DATA = {
1809
1809
  "params": {
1810
1810
  "type": "object",
1811
1811
  "required": false,
1812
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
1812
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
1813
1813
  "properties": [
1814
1814
  {
1815
1815
  "name": "filter",
@@ -2198,7 +2198,7 @@ const CAPABILITY_DATA = {
2198
2198
  "params": {
2199
2199
  "type": "object",
2200
2200
  "required": false,
2201
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
2201
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
2202
2202
  "properties": [
2203
2203
  {
2204
2204
  "name": "filter",
@@ -2794,7 +2794,7 @@ const CAPABILITY_DATA = {
2794
2794
  "params": {
2795
2795
  "type": "object",
2796
2796
  "required": false,
2797
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
2797
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
2798
2798
  "properties": [
2799
2799
  {
2800
2800
  "name": "filter",
@@ -3346,7 +3346,7 @@ const CAPABILITY_DATA = {
3346
3346
  "params": {
3347
3347
  "type": "object",
3348
3348
  "required": false,
3349
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
3349
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
3350
3350
  "properties": [
3351
3351
  {
3352
3352
  "name": "filter",
@@ -4044,7 +4044,7 @@ const CAPABILITY_DATA = {
4044
4044
  "params": {
4045
4045
  "type": "object",
4046
4046
  "required": false,
4047
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
4047
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
4048
4048
  "properties": [
4049
4049
  {
4050
4050
  "name": "filter",
@@ -4557,7 +4557,7 @@ const CAPABILITY_DATA = {
4557
4557
  "params": {
4558
4558
  "type": "object",
4559
4559
  "required": false,
4560
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
4560
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
4561
4561
  "properties": [
4562
4562
  {
4563
4563
  "name": "filter",
@@ -5135,7 +5135,7 @@ const CAPABILITY_DATA = {
5135
5135
  "params": {
5136
5136
  "type": "object",
5137
5137
  "required": false,
5138
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
5138
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
5139
5139
  "properties": [
5140
5140
  {
5141
5141
  "name": "filter",
@@ -5812,7 +5812,7 @@ const CAPABILITY_DATA = {
5812
5812
  "params": {
5813
5813
  "type": "object",
5814
5814
  "required": false,
5815
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
5815
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
5816
5816
  "properties": [
5817
5817
  {
5818
5818
  "name": "filter",
@@ -6418,7 +6418,7 @@ const CAPABILITY_DATA = {
6418
6418
  "params": {
6419
6419
  "type": "object",
6420
6420
  "required": false,
6421
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
6421
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
6422
6422
  "properties": [
6423
6423
  {
6424
6424
  "name": "filter",
@@ -7074,7 +7074,7 @@ const CAPABILITY_DATA = {
7074
7074
  "params": {
7075
7075
  "type": "object",
7076
7076
  "required": false,
7077
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
7077
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
7078
7078
  "properties": [
7079
7079
  {
7080
7080
  "name": "filter",
@@ -7619,7 +7619,7 @@ const CAPABILITY_DATA = {
7619
7619
  "params": {
7620
7620
  "type": "object",
7621
7621
  "required": false,
7622
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
7622
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
7623
7623
  "properties": [
7624
7624
  {
7625
7625
  "name": "filter",
@@ -8292,7 +8292,7 @@ const CAPABILITY_DATA = {
8292
8292
  "params": {
8293
8293
  "type": "object",
8294
8294
  "required": false,
8295
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
8295
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
8296
8296
  "properties": [
8297
8297
  {
8298
8298
  "name": "filter",
@@ -8831,7 +8831,7 @@ const CAPABILITY_DATA = {
8831
8831
  "params": {
8832
8832
  "type": "object",
8833
8833
  "required": false,
8834
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
8834
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
8835
8835
  "properties": [
8836
8836
  {
8837
8837
  "name": "filter",
@@ -9383,7 +9383,7 @@ const CAPABILITY_DATA = {
9383
9383
  "params": {
9384
9384
  "type": "object",
9385
9385
  "required": false,
9386
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
9386
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
9387
9387
  "properties": [
9388
9388
  {
9389
9389
  "name": "filter",
@@ -9964,7 +9964,7 @@ const CAPABILITY_DATA = {
9964
9964
  "params": {
9965
9965
  "type": "object",
9966
9966
  "required": false,
9967
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
9967
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
9968
9968
  "properties": [
9969
9969
  {
9970
9970
  "name": "filter",
@@ -10545,7 +10545,7 @@ const CAPABILITY_DATA = {
10545
10545
  "params": {
10546
10546
  "type": "object",
10547
10547
  "required": false,
10548
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
10548
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
10549
10549
  "properties": [
10550
10550
  {
10551
10551
  "name": "filter",
@@ -11110,7 +11110,7 @@ const CAPABILITY_DATA = {
11110
11110
  "params": {
11111
11111
  "type": "object",
11112
11112
  "required": false,
11113
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
11113
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
11114
11114
  "properties": [
11115
11115
  {
11116
11116
  "name": "filter",
@@ -11474,7 +11474,7 @@ const CAPABILITY_DATA = {
11474
11474
  "params": {
11475
11475
  "type": "object",
11476
11476
  "required": false,
11477
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
11477
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
11478
11478
  "properties": [
11479
11479
  {
11480
11480
  "name": "filter",
@@ -11838,7 +11838,7 @@ const CAPABILITY_DATA = {
11838
11838
  "params": {
11839
11839
  "type": "object",
11840
11840
  "required": false,
11841
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
11841
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
11842
11842
  "properties": [
11843
11843
  {
11844
11844
  "name": "filter",
@@ -12276,7 +12276,7 @@ const CAPABILITY_DATA = {
12276
12276
  "params": {
12277
12277
  "type": "object",
12278
12278
  "required": false,
12279
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
12279
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
12280
12280
  "properties": [
12281
12281
  {
12282
12282
  "name": "filter",
@@ -12909,7 +12909,7 @@ const CAPABILITY_DATA = {
12909
12909
  "params": {
12910
12910
  "type": "object",
12911
12911
  "required": false,
12912
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
12912
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
12913
12913
  "properties": [
12914
12914
  {
12915
12915
  "name": "filter",
@@ -13698,7 +13698,7 @@ const CAPABILITY_DATA = {
13698
13698
  "params": {
13699
13699
  "type": "object",
13700
13700
  "required": false,
13701
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
13701
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
13702
13702
  "properties": [
13703
13703
  {
13704
13704
  "name": "filter",
@@ -14341,7 +14341,7 @@ const CAPABILITY_DATA = {
14341
14341
  "params": {
14342
14342
  "type": "object",
14343
14343
  "required": false,
14344
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
14344
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
14345
14345
  "properties": [
14346
14346
  {
14347
14347
  "name": "filter",
@@ -15016,7 +15016,7 @@ const CAPABILITY_DATA = {
15016
15016
  "params": {
15017
15017
  "type": "object",
15018
15018
  "required": false,
15019
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
15019
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
15020
15020
  "properties": [
15021
15021
  {
15022
15022
  "name": "filter",
@@ -15575,7 +15575,7 @@ const CAPABILITY_DATA = {
15575
15575
  "params": {
15576
15576
  "type": "object",
15577
15577
  "required": false,
15578
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
15578
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
15579
15579
  "properties": [
15580
15580
  {
15581
15581
  "name": "filter",
@@ -16344,7 +16344,7 @@ const CAPABILITY_DATA = {
16344
16344
  "params": {
16345
16345
  "type": "object",
16346
16346
  "required": false,
16347
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
16347
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
16348
16348
  "properties": [
16349
16349
  {
16350
16350
  "name": "filter",
@@ -16973,7 +16973,7 @@ const CAPABILITY_DATA = {
16973
16973
  "params": {
16974
16974
  "type": "object",
16975
16975
  "required": false,
16976
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
16976
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
16977
16977
  "properties": [
16978
16978
  {
16979
16979
  "name": "filter",
@@ -17588,7 +17588,7 @@ const CAPABILITY_DATA = {
17588
17588
  "params": {
17589
17589
  "type": "object",
17590
17590
  "required": false,
17591
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
17591
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
17592
17592
  "properties": [
17593
17593
  {
17594
17594
  "name": "filter",
@@ -18283,7 +18283,7 @@ const CAPABILITY_DATA = {
18283
18283
  "params": {
18284
18284
  "type": "object",
18285
18285
  "required": false,
18286
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
18286
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
18287
18287
  "properties": [
18288
18288
  {
18289
18289
  "name": "filter",
@@ -18715,7 +18715,7 @@ const CAPABILITY_DATA = {
18715
18715
  "params": {
18716
18716
  "type": "object",
18717
18717
  "required": false,
18718
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
18718
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
18719
18719
  "properties": [
18720
18720
  {
18721
18721
  "name": "filter",
@@ -19168,7 +19168,7 @@ const CAPABILITY_DATA = {
19168
19168
  "params": {
19169
19169
  "type": "object",
19170
19170
  "required": false,
19171
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
19171
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
19172
19172
  "properties": [
19173
19173
  {
19174
19174
  "name": "filter",
@@ -19782,7 +19782,7 @@ const CAPABILITY_DATA = {
19782
19782
  "params": {
19783
19783
  "type": "object",
19784
19784
  "required": false,
19785
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
19785
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
19786
19786
  "properties": [
19787
19787
  {
19788
19788
  "name": "filter",
@@ -20274,7 +20274,7 @@ const CAPABILITY_DATA = {
20274
20274
  "params": {
20275
20275
  "type": "object",
20276
20276
  "required": false,
20277
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
20277
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
20278
20278
  "properties": [
20279
20279
  {
20280
20280
  "name": "filter",
@@ -20774,7 +20774,7 @@ const CAPABILITY_DATA = {
20774
20774
  "params": {
20775
20775
  "type": "object",
20776
20776
  "required": false,
20777
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
20777
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
20778
20778
  "properties": [
20779
20779
  {
20780
20780
  "name": "filter",
@@ -21261,7 +21261,7 @@ const CAPABILITY_DATA = {
21261
21261
  "params": {
21262
21262
  "type": "object",
21263
21263
  "required": false,
21264
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
21264
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
21265
21265
  "properties": [
21266
21266
  {
21267
21267
  "name": "filter",
@@ -21961,7 +21961,7 @@ const CAPABILITY_DATA = {
21961
21961
  "params": {
21962
21962
  "type": "object",
21963
21963
  "required": false,
21964
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
21964
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
21965
21965
  "properties": [
21966
21966
  {
21967
21967
  "name": "filter",
@@ -22634,7 +22634,7 @@ const CAPABILITY_DATA = {
22634
22634
  "params": {
22635
22635
  "type": "object",
22636
22636
  "required": false,
22637
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
22637
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
22638
22638
  "properties": [
22639
22639
  {
22640
22640
  "name": "filter",
@@ -23188,7 +23188,7 @@ const CAPABILITY_DATA = {
23188
23188
  "params": {
23189
23189
  "type": "object",
23190
23190
  "required": false,
23191
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
23191
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
23192
23192
  "properties": [
23193
23193
  {
23194
23194
  "name": "filter",
@@ -23692,7 +23692,7 @@ const CAPABILITY_DATA = {
23692
23692
  "params": {
23693
23693
  "type": "object",
23694
23694
  "required": false,
23695
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
23695
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
23696
23696
  "properties": [
23697
23697
  {
23698
23698
  "name": "filter",
@@ -24273,7 +24273,7 @@ const CAPABILITY_DATA = {
24273
24273
  "params": {
24274
24274
  "type": "object",
24275
24275
  "required": false,
24276
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
24276
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
24277
24277
  "properties": [
24278
24278
  {
24279
24279
  "name": "filter",
@@ -24790,7 +24790,7 @@ const CAPABILITY_DATA = {
24790
24790
  "params": {
24791
24791
  "type": "object",
24792
24792
  "required": false,
24793
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
24793
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
24794
24794
  "properties": [
24795
24795
  {
24796
24796
  "name": "filter",
@@ -25404,7 +25404,7 @@ const CAPABILITY_DATA = {
25404
25404
  "params": {
25405
25405
  "type": "object",
25406
25406
  "required": false,
25407
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
25407
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
25408
25408
  "properties": [
25409
25409
  {
25410
25410
  "name": "filter",
@@ -25830,7 +25830,7 @@ const CAPABILITY_DATA = {
25830
25830
  "params": {
25831
25831
  "type": "object",
25832
25832
  "required": false,
25833
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
25833
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
25834
25834
  "properties": [
25835
25835
  {
25836
25836
  "name": "filter",
@@ -26319,7 +26319,7 @@ const CAPABILITY_DATA = {
26319
26319
  "params": {
26320
26320
  "type": "object",
26321
26321
  "required": false,
26322
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
26322
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
26323
26323
  "properties": [
26324
26324
  {
26325
26325
  "name": "filter",
@@ -26762,7 +26762,7 @@ const CAPABILITY_DATA = {
26762
26762
  "params": {
26763
26763
  "type": "object",
26764
26764
  "required": false,
26765
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
26765
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
26766
26766
  "properties": [
26767
26767
  {
26768
26768
  "name": "filter",
@@ -27564,7 +27564,7 @@ const CAPABILITY_DATA = {
27564
27564
  "params": {
27565
27565
  "type": "object",
27566
27566
  "required": false,
27567
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
27567
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
27568
27568
  "properties": [
27569
27569
  {
27570
27570
  "name": "filter",
@@ -28435,7 +28435,7 @@ const CAPABILITY_DATA = {
28435
28435
  "params": {
28436
28436
  "type": "object",
28437
28437
  "required": false,
28438
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
28438
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
28439
28439
  "properties": [
28440
28440
  {
28441
28441
  "name": "filter",
@@ -29304,7 +29304,7 @@ const CAPABILITY_DATA = {
29304
29304
  "params": {
29305
29305
  "type": "object",
29306
29306
  "required": false,
29307
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
29307
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
29308
29308
  "properties": [
29309
29309
  {
29310
29310
  "name": "filter",
@@ -29979,7 +29979,7 @@ const CAPABILITY_DATA = {
29979
29979
  "params": {
29980
29980
  "type": "object",
29981
29981
  "required": false,
29982
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
29982
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
29983
29983
  "properties": [
29984
29984
  {
29985
29985
  "name": "filter",
@@ -30481,7 +30481,7 @@ const CAPABILITY_DATA = {
30481
30481
  "params": {
30482
30482
  "type": "object",
30483
30483
  "required": false,
30484
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
30484
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
30485
30485
  "properties": [
30486
30486
  {
30487
30487
  "name": "filter",
@@ -31066,7 +31066,7 @@ const CAPABILITY_DATA = {
31066
31066
  "params": {
31067
31067
  "type": "object",
31068
31068
  "required": false,
31069
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
31069
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
31070
31070
  "properties": [
31071
31071
  {
31072
31072
  "name": "filter",
@@ -31603,7 +31603,7 @@ const CAPABILITY_DATA = {
31603
31603
  "params": {
31604
31604
  "type": "object",
31605
31605
  "required": false,
31606
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
31606
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
31607
31607
  "properties": [
31608
31608
  {
31609
31609
  "name": "filter",
@@ -32140,7 +32140,7 @@ const CAPABILITY_DATA = {
32140
32140
  "params": {
32141
32141
  "type": "object",
32142
32142
  "required": false,
32143
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
32143
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
32144
32144
  "properties": [
32145
32145
  {
32146
32146
  "name": "filter",
@@ -32677,7 +32677,7 @@ const CAPABILITY_DATA = {
32677
32677
  "params": {
32678
32678
  "type": "object",
32679
32679
  "required": false,
32680
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
32680
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
32681
32681
  "properties": [
32682
32682
  {
32683
32683
  "name": "filter",
@@ -33341,7 +33341,7 @@ const CAPABILITY_DATA = {
33341
33341
  "params": {
33342
33342
  "type": "object",
33343
33343
  "required": false,
33344
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
33344
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
33345
33345
  "properties": [
33346
33346
  {
33347
33347
  "name": "filter",
@@ -33844,7 +33844,7 @@ const CAPABILITY_DATA = {
33844
33844
  "params": {
33845
33845
  "type": "object",
33846
33846
  "required": false,
33847
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
33847
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
33848
33848
  "properties": [
33849
33849
  {
33850
33850
  "name": "filter",
@@ -34483,7 +34483,7 @@ const CAPABILITY_DATA = {
34483
34483
  "params": {
34484
34484
  "type": "object",
34485
34485
  "required": false,
34486
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
34486
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
34487
34487
  "properties": [
34488
34488
  {
34489
34489
  "name": "filter",
@@ -35152,7 +35152,7 @@ const CAPABILITY_DATA = {
35152
35152
  "params": {
35153
35153
  "type": "object",
35154
35154
  "required": false,
35155
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
35155
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
35156
35156
  "properties": [
35157
35157
  {
35158
35158
  "name": "filter",
@@ -35652,7 +35652,7 @@ const CAPABILITY_DATA = {
35652
35652
  "params": {
35653
35653
  "type": "object",
35654
35654
  "required": false,
35655
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
35655
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
35656
35656
  "properties": [
35657
35657
  {
35658
35658
  "name": "filter",
@@ -36075,7 +36075,7 @@ const CAPABILITY_DATA = {
36075
36075
  "params": {
36076
36076
  "type": "object",
36077
36077
  "required": false,
36078
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
36078
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
36079
36079
  "properties": [
36080
36080
  {
36081
36081
  "name": "filter",
@@ -36523,7 +36523,7 @@ const CAPABILITY_DATA = {
36523
36523
  "params": {
36524
36524
  "type": "object",
36525
36525
  "required": false,
36526
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
36526
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
36527
36527
  "properties": [
36528
36528
  {
36529
36529
  "name": "filter",
@@ -37163,7 +37163,7 @@ const CAPABILITY_DATA = {
37163
37163
  "params": {
37164
37164
  "type": "object",
37165
37165
  "required": false,
37166
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
37166
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
37167
37167
  "properties": [
37168
37168
  {
37169
37169
  "name": "filter",
@@ -37714,7 +37714,7 @@ const CAPABILITY_DATA = {
37714
37714
  "params": {
37715
37715
  "type": "object",
37716
37716
  "required": false,
37717
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
37717
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
37718
37718
  "properties": [
37719
37719
  {
37720
37720
  "name": "filter",
@@ -38325,7 +38325,7 @@ const CAPABILITY_DATA = {
38325
38325
  "params": {
38326
38326
  "type": "object",
38327
38327
  "required": false,
38328
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
38328
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
38329
38329
  "properties": [
38330
38330
  {
38331
38331
  "name": "filter",
@@ -38885,7 +38885,7 @@ const CAPABILITY_DATA = {
38885
38885
  "params": {
38886
38886
  "type": "object",
38887
38887
  "required": false,
38888
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
38888
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
38889
38889
  "properties": [
38890
38890
  {
38891
38891
  "name": "filter",
@@ -39367,7 +39367,7 @@ const CAPABILITY_DATA = {
39367
39367
  "params": {
39368
39368
  "type": "object",
39369
39369
  "required": false,
39370
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
39370
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
39371
39371
  "properties": [
39372
39372
  {
39373
39373
  "name": "filter",
@@ -39958,7 +39958,7 @@ const CAPABILITY_DATA = {
39958
39958
  "params": {
39959
39959
  "type": "object",
39960
39960
  "required": false,
39961
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
39961
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
39962
39962
  "properties": [
39963
39963
  {
39964
39964
  "name": "filter",
@@ -40411,7 +40411,7 @@ const CAPABILITY_DATA = {
40411
40411
  "params": {
40412
40412
  "type": "object",
40413
40413
  "required": false,
40414
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
40414
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
40415
40415
  "properties": [
40416
40416
  {
40417
40417
  "name": "filter",
@@ -41034,7 +41034,7 @@ const CAPABILITY_DATA = {
41034
41034
  "params": {
41035
41035
  "type": "object",
41036
41036
  "required": false,
41037
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
41037
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
41038
41038
  "properties": [
41039
41039
  {
41040
41040
  "name": "filter",
@@ -41598,7 +41598,7 @@ const CAPABILITY_DATA = {
41598
41598
  "params": {
41599
41599
  "type": "object",
41600
41600
  "required": false,
41601
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
41601
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
41602
41602
  "properties": [
41603
41603
  {
41604
41604
  "name": "filter",
@@ -42887,7 +42887,7 @@ const CAPABILITY_DATA = {
42887
42887
  "params": {
42888
42888
  "type": "object",
42889
42889
  "required": false,
42890
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
42890
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
42891
42891
  "properties": [
42892
42892
  {
42893
42893
  "name": "filter",
@@ -43838,7 +43838,7 @@ const CAPABILITY_DATA = {
43838
43838
  "params": {
43839
43839
  "type": "object",
43840
43840
  "required": false,
43841
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
43841
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
43842
43842
  "properties": [
43843
43843
  {
43844
43844
  "name": "filter",
@@ -44406,7 +44406,7 @@ const CAPABILITY_DATA = {
44406
44406
  "params": {
44407
44407
  "type": "object",
44408
44408
  "required": false,
44409
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
44409
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
44410
44410
  "properties": [
44411
44411
  {
44412
44412
  "name": "filter",
@@ -45049,7 +45049,7 @@ const CAPABILITY_DATA = {
45049
45049
  "params": {
45050
45050
  "type": "object",
45051
45051
  "required": false,
45052
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
45052
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
45053
45053
  "properties": [
45054
45054
  {
45055
45055
  "name": "filter",
@@ -45567,7 +45567,7 @@ const CAPABILITY_DATA = {
45567
45567
  "params": {
45568
45568
  "type": "object",
45569
45569
  "required": false,
45570
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
45570
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
45571
45571
  "properties": [
45572
45572
  {
45573
45573
  "name": "filter",
@@ -45863,8 +45863,8 @@ const CAPABILITY_DATA = {
45863
45863
  ]
45864
45864
  },
45865
45865
  "A": {
45866
- "planHash": "5a5cfa013a7d398302f69e1c12e260ea29a89942aec743997be1f8cdf60a8df1",
45867
- "builtAt": "2026-11-19T23:12:11.551Z",
45866
+ "planHash": "9521525945b1306da363ce5fb38649e17c5526f2c0842da64c3bec7993e62735",
45867
+ "builtAt": "2026-11-20T05:02:13.666Z",
45868
45868
  "records": [
45869
45869
  {
45870
45870
  "id": "attachment_info.count",
@@ -46145,7 +46145,7 @@ const CAPABILITY_DATA = {
46145
46145
  "params": {
46146
46146
  "type": "object",
46147
46147
  "required": false,
46148
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
46148
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
46149
46149
  "properties": [
46150
46150
  {
46151
46151
  "name": "filter",
@@ -47052,7 +47052,7 @@ const CAPABILITY_DATA = {
47052
47052
  "params": {
47053
47053
  "type": "object",
47054
47054
  "required": false,
47055
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
47055
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
47056
47056
  "properties": [
47057
47057
  {
47058
47058
  "name": "filter",
@@ -48288,7 +48288,7 @@ const CAPABILITY_DATA = {
48288
48288
  "params": {
48289
48289
  "type": "object",
48290
48290
  "required": false,
48291
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
48291
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
48292
48292
  "properties": [
48293
48293
  {
48294
48294
  "name": "filter",
@@ -48882,7 +48882,7 @@ const CAPABILITY_DATA = {
48882
48882
  "params": {
48883
48883
  "type": "object",
48884
48884
  "required": false,
48885
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
48885
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
48886
48886
  "properties": [
48887
48887
  {
48888
48888
  "name": "filter",
@@ -49386,7 +49386,7 @@ const CAPABILITY_DATA = {
49386
49386
  "params": {
49387
49387
  "type": "object",
49388
49388
  "required": false,
49389
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
49389
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
49390
49390
  "properties": [
49391
49391
  {
49392
49392
  "name": "filter",
@@ -50052,7 +50052,7 @@ const CAPABILITY_DATA = {
50052
50052
  "params": {
50053
50053
  "type": "object",
50054
50054
  "required": false,
50055
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
50055
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
50056
50056
  "properties": [
50057
50057
  {
50058
50058
  "name": "filter",
@@ -50501,7 +50501,7 @@ const CAPABILITY_DATA = {
50501
50501
  "params": {
50502
50502
  "type": "object",
50503
50503
  "required": false,
50504
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
50504
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
50505
50505
  "properties": [
50506
50506
  {
50507
50507
  "name": "filter",
@@ -51185,7 +51185,7 @@ const CAPABILITY_DATA = {
51185
51185
  "params": {
51186
51186
  "type": "object",
51187
51187
  "required": false,
51188
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
51188
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
51189
51189
  "properties": [
51190
51190
  {
51191
51191
  "name": "filter",
@@ -51872,7 +51872,7 @@ const CAPABILITY_DATA = {
51872
51872
  "params": {
51873
51873
  "type": "object",
51874
51874
  "required": false,
51875
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
51875
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
51876
51876
  "properties": [
51877
51877
  {
51878
51878
  "name": "filter",
@@ -52547,7 +52547,7 @@ const CAPABILITY_DATA = {
52547
52547
  "params": {
52548
52548
  "type": "object",
52549
52549
  "required": false,
52550
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
52550
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
52551
52551
  "properties": [
52552
52552
  {
52553
52553
  "name": "filter",
@@ -53023,7 +53023,7 @@ const CAPABILITY_DATA = {
53023
53023
  "params": {
53024
53024
  "type": "object",
53025
53025
  "required": false,
53026
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
53026
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
53027
53027
  "properties": [
53028
53028
  {
53029
53029
  "name": "filter",
@@ -53591,7 +53591,7 @@ const CAPABILITY_DATA = {
53591
53591
  "params": {
53592
53592
  "type": "object",
53593
53593
  "required": false,
53594
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
53594
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
53595
53595
  "properties": [
53596
53596
  {
53597
53597
  "name": "filter",
@@ -54136,7 +54136,7 @@ const CAPABILITY_DATA = {
54136
54136
  "params": {
54137
54137
  "type": "object",
54138
54138
  "required": false,
54139
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
54139
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
54140
54140
  "properties": [
54141
54141
  {
54142
54142
  "name": "filter",
@@ -54585,7 +54585,7 @@ const CAPABILITY_DATA = {
54585
54585
  "params": {
54586
54586
  "type": "object",
54587
54587
  "required": false,
54588
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
54588
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
54589
54589
  "properties": [
54590
54590
  {
54591
54591
  "name": "filter",
@@ -55194,7 +55194,7 @@ const CAPABILITY_DATA = {
55194
55194
  "params": {
55195
55195
  "type": "object",
55196
55196
  "required": false,
55197
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
55197
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
55198
55198
  "properties": [
55199
55199
  {
55200
55200
  "name": "filter",
@@ -55696,7 +55696,7 @@ const CAPABILITY_DATA = {
55696
55696
  "params": {
55697
55697
  "type": "object",
55698
55698
  "required": false,
55699
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
55699
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
55700
55700
  "properties": [
55701
55701
  {
55702
55702
  "name": "filter",
@@ -56391,7 +56391,7 @@ const CAPABILITY_DATA = {
56391
56391
  "params": {
56392
56392
  "type": "object",
56393
56393
  "required": false,
56394
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
56394
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
56395
56395
  "properties": [
56396
56396
  {
56397
56397
  "name": "filter",
@@ -56984,7 +56984,8 @@ const CAPABILITY_DATA = {
56984
56984
  "usage": "item: null on the 200 body maps to NotFoundError (vendor not-found signal).",
56985
56985
  "preferredWhen": "Preferred when the numeric id is already known; configuration_items.resolve is for a field value.",
56986
56986
  "related": [
56987
- "configuration_items.resolve"
56987
+ "configuration_items.resolve",
56988
+ "subscriptions.query"
56988
56989
  ],
56989
56990
  "args": "args[0] is the record id (number) Item GET returns the FULL vendor model — the vendor documents no field selection on single-record reads. For a slim record use configuration_items.search (compact summary) or a configuration_items.query with includeFields.",
56990
56991
  "untrustedFields": [
@@ -57391,7 +57392,7 @@ const CAPABILITY_DATA = {
57391
57392
  "opts": {
57392
57393
  "type": "object",
57393
57394
  "required": false,
57394
- "description": "ConfigurationItemProfileOptions — { noteLimit?: number (default 50), relatedLimit?: number (default 50), names?: boolean (default false — opt-in name/label siblings + the top-level `names` section), maxNameReads?: number (default 20 — cold wire name-read budget, lowerable never raised) }.",
57395
+ "description": "ConfigurationItemProfileOptions — { noteLimit?: number (default 50), relatedLimit?: number (default 50), subscriptionLimit?: number (default 50 — the subscriptions part cap; a full page sets capped: true), names?: boolean (default false — opt-in name/label siblings + the top-level `names` section), maxNameReads?: number (default 20 — cold wire name-read budget, lowerable never raised) }.",
57395
57396
  "properties": [
57396
57397
  {
57397
57398
  "name": "noteLimit",
@@ -57405,6 +57406,12 @@ const CAPABILITY_DATA = {
57405
57406
  "optional": true,
57406
57407
  "default": 50
57407
57408
  },
57409
+ {
57410
+ "name": "subscriptionLimit",
57411
+ "type": "number",
57412
+ "optional": true,
57413
+ "default": 50
57414
+ },
57408
57415
  {
57409
57416
  "name": "names",
57410
57417
  "type": "boolean",
@@ -57427,7 +57434,7 @@ const CAPABILITY_DATA = {
57427
57434
  },
57428
57435
  "examples": [
57429
57436
  "await client.configurationItems.profile(1)",
57430
- "Replace the illustrative id with a known configuration item (asset/device) id. The profile reads the item and its related records; a missing item returns NOT_FOUND. Product-association rows are returned from the scoped query, without inferring tenant inventory from a product count."
57437
+ "Replace the illustrative id with a known configuration item (asset/device) id. The profile reads the item and its related records, including its product and subscriptions; a missing item returns NOT_FOUND. Product-association and subscription rows are returned from scoped queries, without inferring tenant inventory from a count."
57431
57438
  ],
57432
57439
  "permissions": "unknown",
57433
57440
  "pagination": {
@@ -57441,9 +57448,9 @@ const CAPABILITY_DATA = {
57441
57448
  "set": 13
57442
57449
  },
57443
57450
  "metadata": {
57444
- "purpose": "Get the full profile of one configuration item in a single call: the item, its type and category records, its notes, related items and billing-product associations.",
57445
- "usage": "args[0] is the configuration item id (number); args[1] (optional) is the options object { noteLimit?: number (default 50), relatedLimit?: number (default 50), names?: boolean (default false — ON adds the opt-in name sibling on the ci row (closed scope: the contract name — the CI row has no issue-listed picklist fields, so no labels) + the top-level `names` section; the sibling rides the existing default-on caches (NO new cache)), maxNameReads?: number (default 20 — the budget of COLD wire name reads, lowerable never raised; over-budget ids serve name: null and the names section sets `truncated: true`, never an error) }. Wire: <=6 reads (get + type + category + 3 item-scoped queries; a dangling type/category FK skips its read). With names: true, N added name reads (N = the distinct uncached FK ids, capped by maxNameReads). productAssociations contains the scoped query rows and capped flag, with blind: false on a successful read; a product create requirement does not imply that associations are unreadable. A missing configuration item returns NOT_FOUND.",
57446
- "preferredWhen": "The single read for 'what is this configuration item' — replaces 5-6 raw reads + describes with one call.",
57451
+ "purpose": "Get the full profile of one configuration item in a single call: the item, its type, category and product records, its notes, related items, billing-product associations and subscriptions (software/licence terms keyed by configurationItemID).",
57452
+ "usage": "args[0] is the configuration item id (number); args[1] (optional) is the options object { noteLimit?: number (default 50), relatedLimit?: number (default 50), subscriptionLimit?: number (default 50), names?: boolean (default false — ON adds the opt-in name siblings on the ci row (closed scope: contractIDName, companyIDName, productIDName and vendorIDName, where vendorID names a company — the CI row has no issue-listed picklist fields, so no labels) + the top-level `names` section; the siblings ride the existing default-on caches (NO new cache)), maxNameReads?: number (default 20 — the budget of COLD wire name reads, lowerable never raised; over-budget ids serve name: null and the names section sets `truncated: true`, never an error) }. Wire: <=8 reads (get + type + category + product + 4 item-scoped queries; a missing or dangling type/category/product FK skips its read and serves null with parts.<part> 'absent'; productID 0 is never read). The product and subscriptions parts degrade on a typed vendor fault (e.g. an API user without read access to Products or Subscriptions): product null with parts.product 'unavailable', subscriptions { items: [], capped: false } with parts.subscriptions 'unavailable', and the fault code under unavailable.<part> (present only when a part degraded); a fault on any other part rejects the profile. subscriptions is the Subscriptions query filtered configurationItemID eq the item id, capped at subscriptionLimit (capped: true on a full page — more may exist). With names: true, N added name reads (N = the distinct uncached FK ids, capped by maxNameReads; productIDName comes from the product part's outcome, with no second read). productAssociations contains the scoped query rows and capped flag, with blind: false on a successful read; a product create requirement does not imply that associations are unreadable. A missing configuration item returns NOT_FOUND.",
57453
+ "preferredWhen": "The single read for 'what is this configuration item' — replaces 7-8 raw reads + describes with one call, including the item's subscription terms (start/end dates, vendor, period price/cost).",
57447
57454
  "related": [
57448
57455
  "configuration_items.get",
57449
57456
  "configuration_items.findBySerial",
@@ -57451,9 +57458,11 @@ const CAPABILITY_DATA = {
57451
57458
  "configuration_item_related_items.query",
57452
57459
  "configuration_item_billing_product_associations.query",
57453
57460
  "configuration_item_types.get",
57454
- "configuration_item_categories.get"
57461
+ "configuration_item_categories.get",
57462
+ "products.get",
57463
+ "subscriptions.query"
57455
57464
  ],
57456
- "args": "args[0] is the configuration item id (number); args[1] (optional) is the options object { noteLimit?: number (default 50), relatedLimit?: number (default 50), names?: boolean (default false — the opt-in contract-name sibling on the ci row + the `names` section), maxNameReads?: number (default 20, lowerable never raised — the budget of cold wire name reads) }. <=6 bounded reads inside, zero mutating; product associations are returned from the scoped vendor query without a static inventory or visibility assumption.",
57465
+ "args": "args[0] is the configuration item id (number); args[1] (optional) is the options object { noteLimit?: number (default 50), relatedLimit?: number (default 50), subscriptionLimit?: number (default 50), names?: boolean (default false — the opt-in contract/company/product/vendor name siblings on the ci row + the `names` section), maxNameReads?: number (default 20, lowerable never raised — the budget of cold wire name reads) }. <=8 bounded reads inside, zero mutating; product associations and subscriptions are returned from scoped vendor queries without a static inventory or visibility assumption; the product and subscriptions parts degrade to 'unavailable' (with the fault code) on a typed vendor fault.",
57457
57466
  "untrustedFields": [
57458
57467
  "UDF1",
57459
57468
  "UDF10",
@@ -57475,14 +57484,21 @@ const CAPABILITY_DATA = {
57475
57484
  "UDF7",
57476
57485
  "UDF8",
57477
57486
  "UDF9",
57487
+ "companyIDName",
57478
57488
  "contractIDName",
57479
57489
  "description",
57490
+ "link",
57480
57491
  "location",
57492
+ "manufacturerName",
57493
+ "manufacturerProductName",
57481
57494
  "name",
57482
57495
  "nickname",
57483
57496
  "notes",
57497
+ "productIDName",
57484
57498
  "referenceTitle",
57485
- "title"
57499
+ "subscriptionName",
57500
+ "title",
57501
+ "vendorIDName"
57486
57502
  ]
57487
57503
  },
57488
57504
  "compact": "ConfigurationItemProfile",
@@ -57503,7 +57519,7 @@ const CAPABILITY_DATA = {
57503
57519
  "params": {
57504
57520
  "type": "object",
57505
57521
  "required": false,
57506
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
57522
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
57507
57523
  "properties": [
57508
57524
  {
57509
57525
  "name": "filter",
@@ -57591,7 +57607,8 @@ const CAPABILITY_DATA = {
57591
57607
  "usage": "Server-side pagination: one call walks the vendor cursor internally and preserves the initial HTTP method until the limit is met or the cursor is null. By default, each POST carries the same QueryModel body (IncludeFields, MaxRecords and the original filter). The vendor honours IncludeFields on POST continuations and uses its own continuation page size. With `via: 'get'`, the SDK follows the GET cursor with the original search parameters. Cursors must stay on the configured origin; 1000-page guard (hasMore stays true at the guard). queryAll is the unbounded raw walk — only when you need the ENTIRE set and accept a large payload; default to the bounded query op. noteChars: N (integer >= 1) is an opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose — the same set the server wraps): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result (zero wire — response-side only). latest: N returns the N highest-id rows (id-descending): the vendor has no sort parameter — this is the only ordering channel, and a date window in `filter` narrows the cost; count-anchored — C ≤ 2500 → one count + a ≤6-page walk of the matching set, C > 2500 → count probes + an id-threshold window (≤64 probes; LATEST_THRESHOLD_UNRESOLVED on budget exhaustion), a blind count → a bounded walk with the blind-surface stamp (LATEST_WINDOW_INCOMPLETE on a truncated walk). Explicit walk.maxRows independently sets the TOTAL returned-row cap (including export-sized walks); the ordinary limit is still validated in 1..2500 and only bounds the internal page size for this form. hasMore and truncated remain true at the page guard or while a cursor or unserved fetched rows remain. Self-resource prefixes are accepted in filter fields and includeFields; dotted UDF names are always refused before metadata or vendor requests, never rewritten to another tenant UDF.",
57592
57608
  "preferredWhen": "Default for filtered reads (the bounded query op). queryAll is the unbounded raw walk — only when you need the ENTIRE set and accept a large payload; default to the bounded query op. Prefer an includeFields subset over the full model (field names via operations.fieldCatalog). latest: N is the latest-N rows (id-descending) when the newest rows are the goal (the vendor has no sort — count-anchored, docs/API.md §4). Prefer a hand-listed includeFields for common narrow reads (removed the closed fieldsPreset presets — the v1 billing_join field list for billing_items.query, if you need it, is: companyID, ticketID, projectID, timeEntryID, contractServiceID, contractID, billingCodeID, expenseItemID, itemDate, quantity, rate, extendedPrice, totalAmount, nonBillable); noteChars: N caps the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose) to N chars each with the explicit … [+N chars truncated] marker (opt-in; the raw result when absent).",
57593
57609
  "related": [
57594
- "configuration_items.resolve"
57610
+ "configuration_items.resolve",
57611
+ "subscriptions.query"
57595
57612
  ],
57596
57613
  "args": "args[0] is a params object: { filter?: FilterCondition[], limit?: total cap (default 500, max 2500; server-side pagination), latest?: number (1-2500 — the N highest-id rows; the vendor has no sort parameter, so this is a count-anchored mechanism — see docs/API.md §4; coexists with limit only when limit >= latest), noteChars?: number (integer >= 1 — opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result) maxRecords?: page size — raw SDK `queryAll`/`query` walks only (this bounded query ignores it; the invoke bag refuses it — pass limit), includeFields?: string[], fieldCheck?: boolean (default true — the pre-request field-name check on includeFields names and filter condition fields; false disables it — consumed by the invoke router, never sent to the vendor), via?: \"post\" | \"get\" (the wire method; the vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result’s scopeNote says so), walk?: { maxRows: positive integer (D6: the walk’s TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result’s truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) } }. Filter grammar: 13 operators: eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn (in/notIn: array or '1,2,3') + and/or grouping (no not-group). Field names in vendor API casing — get them via operations.fieldCatalog. Caps: 500 OR-conditions / 1 UDF / 8000 chars. For a simple text lookup prefer configuration_items.search. By default the FULL vendor model is returned (all fields). To slim the payload for context savings pass includeFields with the exact vendor field names you need (discover them via operations.fieldCatalog); only those fields + id come back. The subset persists across the internal page walk.",
57597
57614
  "untrustedFields": [
@@ -58282,7 +58299,7 @@ const CAPABILITY_DATA = {
58282
58299
  "params": {
58283
58300
  "type": "object",
58284
58301
  "required": false,
58285
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
58302
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
58286
58303
  "properties": [
58287
58304
  {
58288
58305
  "name": "filter",
@@ -59558,7 +59575,7 @@ const CAPABILITY_DATA = {
59558
59575
  "params": {
59559
59576
  "type": "object",
59560
59577
  "required": false,
59561
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
59578
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
59562
59579
  "properties": [
59563
59580
  {
59564
59581
  "name": "filter",
@@ -60353,7 +60370,7 @@ const CAPABILITY_DATA = {
60353
60370
  "params": {
60354
60371
  "type": "object",
60355
60372
  "required": false,
60356
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
60373
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
60357
60374
  "properties": [
60358
60375
  {
60359
60376
  "name": "filter",
@@ -60908,7 +60925,7 @@ const CAPABILITY_DATA = {
60908
60925
  "params": {
60909
60926
  "type": "object",
60910
60927
  "required": false,
60911
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
60928
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
60912
60929
  "properties": [
60913
60930
  {
60914
60931
  "name": "filter",
@@ -61454,7 +61471,7 @@ const CAPABILITY_DATA = {
61454
61471
  "params": {
61455
61472
  "type": "object",
61456
61473
  "required": false,
61457
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
61474
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
61458
61475
  "properties": [
61459
61476
  {
61460
61477
  "name": "filter",
@@ -62086,7 +62103,7 @@ const CAPABILITY_DATA = {
62086
62103
  "params": {
62087
62104
  "type": "object",
62088
62105
  "required": false,
62089
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
62106
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
62090
62107
  "properties": [
62091
62108
  {
62092
62109
  "name": "filter",
@@ -62768,7 +62785,7 @@ const CAPABILITY_DATA = {
62768
62785
  "params": {
62769
62786
  "type": "object",
62770
62787
  "required": false,
62771
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
62788
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
62772
62789
  "properties": [
62773
62790
  {
62774
62791
  "name": "filter",
@@ -63305,7 +63322,7 @@ const CAPABILITY_DATA = {
63305
63322
  "params": {
63306
63323
  "type": "object",
63307
63324
  "required": false,
63308
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
63325
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
63309
63326
  "properties": [
63310
63327
  {
63311
63328
  "name": "filter",
@@ -63815,7 +63832,7 @@ const CAPABILITY_DATA = {
63815
63832
  "params": {
63816
63833
  "type": "object",
63817
63834
  "required": false,
63818
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
63835
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
63819
63836
  "properties": [
63820
63837
  {
63821
63838
  "name": "filter",
@@ -64424,7 +64441,7 @@ const CAPABILITY_DATA = {
64424
64441
  "params": {
64425
64442
  "type": "object",
64426
64443
  "required": false,
64427
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
64444
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
64428
64445
  "properties": [
64429
64446
  {
64430
64447
  "name": "filter",
@@ -64873,7 +64890,7 @@ const CAPABILITY_DATA = {
64873
64890
  "params": {
64874
64891
  "type": "object",
64875
64892
  "required": false,
64876
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
64893
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
64877
64894
  "properties": [
64878
64895
  {
64879
64896
  "name": "filter",
@@ -65431,7 +65448,7 @@ const CAPABILITY_DATA = {
65431
65448
  "params": {
65432
65449
  "type": "object",
65433
65450
  "required": false,
65434
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
65451
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
65435
65452
  "properties": [
65436
65453
  {
65437
65454
  "name": "filter",
@@ -65977,7 +65994,7 @@ const CAPABILITY_DATA = {
65977
65994
  "params": {
65978
65995
  "type": "object",
65979
65996
  "required": false,
65980
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
65997
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
65981
65998
  "properties": [
65982
65999
  {
65983
66000
  "name": "filter",
@@ -66523,7 +66540,7 @@ const CAPABILITY_DATA = {
66523
66540
  "params": {
66524
66541
  "type": "object",
66525
66542
  "required": false,
66526
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
66543
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
66527
66544
  "properties": [
66528
66545
  {
66529
66546
  "name": "filter",
@@ -67509,7 +67526,7 @@ const CAPABILITY_DATA = {
67509
67526
  "params": {
67510
67527
  "type": "object",
67511
67528
  "required": false,
67512
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
67529
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
67513
67530
  "properties": [
67514
67531
  {
67515
67532
  "name": "filter",
@@ -68019,7 +68036,7 @@ const CAPABILITY_DATA = {
68019
68036
  "params": {
68020
68037
  "type": "object",
68021
68038
  "required": false,
68022
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
68039
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
68023
68040
  "properties": [
68024
68041
  {
68025
68042
  "name": "filter",
@@ -68519,7 +68536,7 @@ const CAPABILITY_DATA = {
68519
68536
  "params": {
68520
68537
  "type": "object",
68521
68538
  "required": false,
68522
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
68539
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
68523
68540
  "properties": [
68524
68541
  {
68525
68542
  "name": "filter",
@@ -69029,7 +69046,7 @@ const CAPABILITY_DATA = {
69029
69046
  "params": {
69030
69047
  "type": "object",
69031
69048
  "required": false,
69032
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
69049
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
69033
69050
  "properties": [
69034
69051
  {
69035
69052
  "name": "filter",
@@ -69587,7 +69604,7 @@ const CAPABILITY_DATA = {
69587
69604
  "params": {
69588
69605
  "type": "object",
69589
69606
  "required": false,
69590
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
69607
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
69591
69608
  "properties": [
69592
69609
  {
69593
69610
  "name": "filter",
@@ -70779,7 +70796,7 @@ const CAPABILITY_DATA = {
70779
70796
  "params": {
70780
70797
  "type": "object",
70781
70798
  "required": false,
70782
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
70799
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
70783
70800
  "properties": [
70784
70801
  {
70785
70802
  "name": "filter",
@@ -71602,7 +71619,7 @@ const CAPABILITY_DATA = {
71602
71619
  "params": {
71603
71620
  "type": "object",
71604
71621
  "required": false,
71605
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
71622
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
71606
71623
  "properties": [
71607
71624
  {
71608
71625
  "name": "filter",
@@ -72338,7 +72355,7 @@ const CAPABILITY_DATA = {
72338
72355
  "params": {
72339
72356
  "type": "object",
72340
72357
  "required": false,
72341
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
72358
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
72342
72359
  "properties": [
72343
72360
  {
72344
72361
  "name": "filter",
@@ -73104,7 +73121,7 @@ const CAPABILITY_DATA = {
73104
73121
  "params": {
73105
73122
  "type": "object",
73106
73123
  "required": false,
73107
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
73124
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
73108
73125
  "properties": [
73109
73126
  {
73110
73127
  "name": "filter",
@@ -73620,7 +73637,7 @@ const CAPABILITY_DATA = {
73620
73637
  "params": {
73621
73638
  "type": "object",
73622
73639
  "required": false,
73623
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
73640
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
73624
73641
  "properties": [
73625
73642
  {
73626
73643
  "name": "filter",
@@ -74229,7 +74246,7 @@ const CAPABILITY_DATA = {
74229
74246
  "params": {
74230
74247
  "type": "object",
74231
74248
  "required": false,
74232
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
74249
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
74233
74250
  "properties": [
74234
74251
  {
74235
74252
  "name": "filter",
@@ -74748,7 +74765,7 @@ const CAPABILITY_DATA = {
74748
74765
  "params": {
74749
74766
  "type": "object",
74750
74767
  "required": false,
74751
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
74768
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
74752
74769
  "properties": [
74753
74770
  {
74754
74771
  "name": "filter",
@@ -75430,7 +75447,7 @@ const CAPABILITY_DATA = {
75430
75447
  "params": {
75431
75448
  "type": "object",
75432
75449
  "required": false,
75433
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
75450
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
75434
75451
  "properties": [
75435
75452
  {
75436
75453
  "name": "filter",
@@ -75879,7 +75896,7 @@ const CAPABILITY_DATA = {
75879
75896
  "params": {
75880
75897
  "type": "object",
75881
75898
  "required": false,
75882
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
75899
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
75883
75900
  "properties": [
75884
75901
  {
75885
75902
  "name": "filter",
@@ -76698,7 +76715,7 @@ const CAPABILITY_DATA = {
76698
76715
  "params": {
76699
76716
  "type": "object",
76700
76717
  "required": false,
76701
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
76718
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
76702
76719
  "properties": [
76703
76720
  {
76704
76721
  "name": "filter",
@@ -77484,7 +77501,7 @@ const CAPABILITY_DATA = {
77484
77501
  "params": {
77485
77502
  "type": "object",
77486
77503
  "required": false,
77487
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
77504
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
77488
77505
  "properties": [
77489
77506
  {
77490
77507
  "name": "filter",
@@ -78101,7 +78118,7 @@ const CAPABILITY_DATA = {
78101
78118
  "params": {
78102
78119
  "type": "object",
78103
78120
  "required": false,
78104
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
78121
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
78105
78122
  "properties": [
78106
78123
  {
78107
78124
  "name": "filter",
@@ -78669,7 +78686,7 @@ const CAPABILITY_DATA = {
78669
78686
  "params": {
78670
78687
  "type": "object",
78671
78688
  "required": false,
78672
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
78689
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
78673
78690
  "properties": [
78674
78691
  {
78675
78692
  "name": "filter",
@@ -79575,7 +79592,7 @@ const CAPABILITY_DATA = {
79575
79592
  "params": {
79576
79593
  "type": "object",
79577
79594
  "required": false,
79578
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
79595
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
79579
79596
  "properties": [
79580
79597
  {
79581
79598
  "name": "filter",
@@ -80228,7 +80245,7 @@ const CAPABILITY_DATA = {
80228
80245
  "params": {
80229
80246
  "type": "object",
80230
80247
  "required": false,
80231
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
80248
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
80232
80249
  "properties": [
80233
80250
  {
80234
80251
  "name": "filter",
@@ -80704,7 +80721,7 @@ const CAPABILITY_DATA = {
80704
80721
  "params": {
80705
80722
  "type": "object",
80706
80723
  "required": false,
80707
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
80724
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
80708
80725
  "properties": [
80709
80726
  {
80710
80727
  "name": "filter",
@@ -81151,7 +81168,7 @@ const CAPABILITY_DATA = {
81151
81168
  "params": {
81152
81169
  "type": "object",
81153
81170
  "required": false,
81154
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
81171
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
81155
81172
  "properties": [
81156
81173
  {
81157
81174
  "name": "filter",
@@ -81686,7 +81703,7 @@ const CAPABILITY_DATA = {
81686
81703
  "params": {
81687
81704
  "type": "object",
81688
81705
  "required": false,
81689
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
81706
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
81690
81707
  "properties": [
81691
81708
  {
81692
81709
  "name": "filter",
@@ -82215,7 +82232,7 @@ const CAPABILITY_DATA = {
82215
82232
  "params": {
82216
82233
  "type": "object",
82217
82234
  "required": false,
82218
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
82235
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
82219
82236
  "properties": [
82220
82237
  {
82221
82238
  "name": "filter",
@@ -82846,7 +82863,7 @@ const CAPABILITY_DATA = {
82846
82863
  "params": {
82847
82864
  "type": "object",
82848
82865
  "required": false,
82849
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
82866
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
82850
82867
  "properties": [
82851
82868
  {
82852
82869
  "name": "filter",
@@ -83604,7 +83621,7 @@ const CAPABILITY_DATA = {
83604
83621
  "params": {
83605
83622
  "type": "object",
83606
83623
  "required": false,
83607
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
83624
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
83608
83625
  "properties": [
83609
83626
  {
83610
83627
  "name": "filter",
@@ -84141,7 +84158,7 @@ const CAPABILITY_DATA = {
84141
84158
  "params": {
84142
84159
  "type": "object",
84143
84160
  "required": false,
84144
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
84161
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
84145
84162
  "properties": [
84146
84163
  {
84147
84164
  "name": "filter",
@@ -84755,7 +84772,7 @@ const CAPABILITY_DATA = {
84755
84772
  "params": {
84756
84773
  "type": "object",
84757
84774
  "required": false,
84758
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
84775
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
84759
84776
  "properties": [
84760
84777
  {
84761
84778
  "name": "filter",
@@ -85319,7 +85336,7 @@ const CAPABILITY_DATA = {
85319
85336
  "params": {
85320
85337
  "type": "object",
85321
85338
  "required": false,
85322
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
85339
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
85323
85340
  "properties": [
85324
85341
  {
85325
85342
  "name": "filter",
@@ -85966,7 +85983,7 @@ const CAPABILITY_DATA = {
85966
85983
  "params": {
85967
85984
  "type": "object",
85968
85985
  "required": false,
85969
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
85986
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
85970
85987
  "properties": [
85971
85988
  {
85972
85989
  "name": "filter",
@@ -86679,7 +86696,7 @@ const CAPABILITY_DATA = {
86679
86696
  "params": {
86680
86697
  "type": "object",
86681
86698
  "required": false,
86682
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
86699
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
86683
86700
  "properties": [
86684
86701
  {
86685
86702
  "name": "filter",
@@ -87567,7 +87584,7 @@ const CAPABILITY_DATA = {
87567
87584
  "params": {
87568
87585
  "type": "object",
87569
87586
  "required": false,
87570
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
87587
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
87571
87588
  "properties": [
87572
87589
  {
87573
87590
  "name": "filter",
@@ -88176,7 +88193,7 @@ const CAPABILITY_DATA = {
88176
88193
  "params": {
88177
88194
  "type": "object",
88178
88195
  "required": false,
88179
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
88196
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
88180
88197
  "properties": [
88181
88198
  {
88182
88199
  "name": "filter",
@@ -88734,7 +88751,7 @@ const CAPABILITY_DATA = {
88734
88751
  "params": {
88735
88752
  "type": "object",
88736
88753
  "required": false,
88737
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
88754
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
88738
88755
  "properties": [
88739
88756
  {
88740
88757
  "name": "filter",
@@ -90197,7 +90214,7 @@ const CAPABILITY_DATA = {
90197
90214
  "params": {
90198
90215
  "type": "object",
90199
90216
  "required": false,
90200
- "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/*ResourceID (contractIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
90217
+ "description": "The query params object (CLOSED — unknown keys are refused pre-request with a typed error naming the key + the valid keys): { filter?: FilterCondition[], limit?: number (total cap, default 500, max 2500 — NOT a page size), latest?: number (1-2500; the N highest-id rows — the vendor has no sort; count-probe + id-threshold semantics, see docs/API.md §4; the same-process latest-walk count probes are briefly cached per client — a 30 s TTL, cleared on a write through this client (a probe in flight across the write is discarded by the cache's invalidation epoch); the plain count op and aggregate stay fresh; limit and latest may coexist only when limit >= latest), noteChars?: number (integer >= 1; opt-in per-field char cap on the untrusted-class free-text fields of the resource (notes/descriptions/UDF prose): each capped value = the first N chars + the explicit marker … [+N chars truncated] (N = chars removed), applied at result-build BEFORE the server-side untrusted wrapper; absent = the raw, untruncated result), includeFields?: string[], fieldCheck?: boolean (default true: every includeFields NAME and every filter condition FIELD NAME is checked pre-request against the live field set (the vendored spec when that read faults — the resource's plan-pinned live-only fields, metadata.liveOnlyFields, are exempt from the vendored refusal: the live entity carries them and the vendor model omits them (e.g. time_entries postedDate/postedOnTime) — a truly unknown name is still refused); a refusal costs zero QUERY wire — the query POST and the count probe that anchors latest are never issued, a COLD fieldCache performs one authority metadata GET first (zero on a warm one); the one udf:true condition per query is exempt (a tenant UDF name is absent from the vendored spec by definition); false disables both name checks, never the shape/grammar checks; the count verb's bag is name-checked the same way and carries the same fieldCheck opt-out; a direct queryBounded/queryAll call is not checked (no bag contract there)), names?: boolean (absent = the client's labels option — OFF by default (the label pass costs a field-catalog read + up to 20 cold name reads per call; the MCP server (iit-mcp-autotask) enables the client labels option when it re-pins to this version (downstream)); true adds the row label siblings: picklist <field>Label (statusLabel), <fk>Name for companyID/contractID/ticketID/roleID/billingCodeID/productID/vendorID (a company)/configurationItemCategoryID/*ResourceID (contractIDName, productIDName, assignedResourceIDName) — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); a names summary appears only when something was left unlabelled (the 20 uncached name-read cap or an unresolved picklist value); false = raw ids, ~30% smaller on a full tickets row), via?: \"post\" | \"get\" (vendor GET has a 2048-char cap), acknowledgeUnscoped?: boolean (query verb only; a walk expected to read over 500 rows with no id scope in the filter is guarded — one pre-flight <resource>.count, and a count over 500 refuses with the typed SCOPE_REQUIRED before the walk; true walks the unscoped set anyway; a count that fails or is blind lets the walk run and the result's scopeNote says so), walk?: object (walk.maxRows: positive integer — (D6): the walk's TOTAL row cap — serves at most maxRows rows in total; the page size stays min(500, limit), so one vendor page may fetch more rows than maxRows and the surplus is sliced off (fetched, not served); the result's truncated flag — false only when the cursor is exhausted, true iff more rows exist past the cap (so an exact-multiple count with a live next page reports true); not combinable with latest) }. Filter grammar: 13 operators (eq, noteq, gt, gte, lt, lte, beginsWith, endsWith, contains, exist, notExist, in, notIn; in/notIn: array or \"1,2,3\") + and/or grouping (no not-group); field names in vendor API casing (via operations.fieldCatalog); caps: 500 OR-conditions / 1 UDF / 8000 chars.",
90201
90218
  "properties": [
90202
90219
  {
90203
90220
  "name": "filter",
@@ -91031,8 +91048,8 @@ const CAPABILITY_DATA = {
91031
91048
  ]
91032
91049
  },
91033
91050
  "operations": {
91034
- "planHash": "5a5cfa013a7d398302f69e1c12e260ea29a89942aec743997be1f8cdf60a8df1",
91035
- "builtAt": "2026-11-19T23:12:11.551Z",
91051
+ "planHash": "9521525945b1306da363ce5fb38649e17c5526f2c0842da64c3bec7993e62735",
91052
+ "builtAt": "2026-11-20T05:02:13.666Z",
91036
91053
  "records": [
91037
91054
  {
91038
91055
  "id": "operations.aggregate",
@@ -91059,7 +91076,7 @@ const CAPABILITY_DATA = {
91059
91076
  "opts": {
91060
91077
  "type": "object",
91061
91078
  "required": false,
91062
- "description": "AggregateOptions — { groupBy?: string[] (vendor field names), sum?: string[] (numeric vendor fields), top?: number (default 25, clamped 1..100 — the group-row cap), limit?: number (total-record cap for the walk modes, default 2500 = the house BOUNDED_QUERY_MAX_LIMIT — the hard max), names?: boolean (absent = the client's labels option — OFF by default; true — group keys gain label siblings, e.g. key.ticketIDName, key.statusLabel — null, never missing, when the field is keyed with a null value, a 0 id or an unresolved value (a picklist 0 is labelled when the picklist defines it); false = raw ids), acknowledgeUnscoped?: boolean (a group/sum with NO filter — absent or empty — walks the whole tenant table and is refused PRE-REQUEST with the typed SCOPE_REQUIRED (zero wire) unless this is true; the count mode is unaffected — one vendor count, no walk) }. NEITHER groupBy nor sum = COUNT mode (one vendor /query/count call — no walk; a 0 on a plan-pinned blind surface is marked blind: true with the sentinel — never a silent absence; a filtered 0 on an unfiltered-blind surface is an unstamped result under current API-user visibility, not proof of tenant-wide absence). The walk modes fold client-side: hasMore/totalScanned carry the truncation honesty (no silent partial sums) and a mid-walk RATE_LIMITED propagates (the partial fold is discarded).",
91079
+ "description": "AggregateOptions — { groupBy?: string[] (vendor field names), sum?: string[] (numeric vendor fields), top?: number (default 25, clamped 1..100 — the group-row cap), limit?: number (total-record cap for the walk modes, default 2500 = the house BOUNDED_QUERY_MAX_LIMIT — the hard max), names?: boolean (absent = the client's labels option — OFF by default; true — group keys gain label siblings, e.g. key.ticketIDName, key.statusLabel — null, never missing, when the field is keyed with a null value, a 0 id (except companyID/vendorID 0, which name the tenant owner's own company) or an unresolved value (a picklist 0 is labelled when the picklist defines it); false = raw ids), acknowledgeUnscoped?: boolean (a group/sum with NO filter — absent or empty — walks the whole tenant table and is refused PRE-REQUEST with the typed SCOPE_REQUIRED (zero wire) unless this is true; the count mode is unaffected — one vendor count, no walk) }. NEITHER groupBy nor sum = COUNT mode (one vendor /query/count call — no walk; a 0 on a plan-pinned blind surface is marked blind: true with the sentinel — never a silent absence; a filtered 0 on an unfiltered-blind surface is an unstamped result under current API-user visibility, not proof of tenant-wide absence). The walk modes fold client-side: hasMore/totalScanned carry the truncation honesty (no silent partial sums) and a mid-walk RATE_LIMITED propagates (the partial fold is discarded).",
91063
91080
  "properties": [
91064
91081
  {
91065
91082
  "name": "groupBy",