@wowok/agent-mcp 3.2.6 → 3.2.9

package/dist/schemas/onchain_operations.schema.json

@@ -97,7 +97,7 @@
               "description": "Two ways to specify an object: (1) STRING - Reference an EXISTING object by name or ID; (2) OBJECT - CREATE a NEW object with full configuration including name, tags, description, token type for payments, and Permission-based access control. The object format is used when you need both token specification AND permission management."
             },
             "order_new": {
-              "description": "Create new order.",
+              "description": "Create new order. RETRY DISCIPLINE: the order creation is atomic on-chain but NOT in observability — a transport drop after submit leaves the commit state UNKNOWN, and a blind resubmit creates a second order with a second full escrow (live mainnet incident). On any failure/timeout, verify whether an order with this exact signature (service + buyer + items) already landed — via the in-flight marker in the failure envelope, the sender's recent on-chain objects, or onchain_events — and adopt it if present; resubmit only if confirmed absent.",
               "type": "object",
               "properties": {
                 "buy": {
@@ -139,6 +139,7 @@
                       "description": "List of products or services to purchase"
                     },
                     "total_pay": {
+                      "description": "Actual payment amount. FORMAT: {balance: <amount_in_smallest_unit>} or {coin: <coin_object_id>}. The token type and precision are determined by the Service object's type_parameter (the generic type set when the Service was created). For WOW (9 decimals): {balance: 1000000000} = 1 WOW. OPTIONAL without a discount (M-K17): omit it and the handler derives the EXACT total Σ(sale price × quantity) from the Service's on-chain sales list and discloses the derived value — no transcription, no drift. REQUIRED when 'discount' is set (the discounted total depends on the coupon's live value: run query_toolkit query_type='service_quote' and pass total_pay.balance = the quoted final_total). An explicit value is always honored as-is. AMOUNT GATE (M-K11-a): a total that EXACTLY equals Σ published sale price × quantity (catalog-priced buy, in the Service's own payment coin — WOW or bridged USDT/USDC/ETH/BTC) bypasses the harness pre-sign amount gate; any other amount above the threshold is blocked until WOWOK_AMOUNT_THRESHOLD is raised after explicit user confirmation.",
                       "anyOf": [
                         {
                           "type": "object",
@@ -174,8 +175,7 @@
                           ],
                           "additionalProperties": false
                         }
-                      ],
-                      "description": "Actual payment amount. FORMAT: {balance: <amount_in_smallest_unit>} or {coin: <coin_object_id>}. The token type and precision are determined by the Service object's type_parameter (the generic type set when the Service was created). For WOW (9 decimals): {balance: 1000000000} = 1 WOW. "
+                      ]
                     },
                     "discount": {
                       "description": "Discount object ID or name",
@@ -191,8 +191,7 @@
                     }
                   },
                   "required": [
-                    "items",
-                    "total_pay"
+                    "items"
                   ],
                   "additionalProperties": false
                 },
@@ -231,7 +230,7 @@
                   "additionalProperties": false
                 },
                 "order_required_info": {
-                  "description": "Contact object ID or WTS Proof object, RECORDED on the Order as evidence that the order holder delivered the private information the Service declares in customer_required (e.g. phone/email/shipping_address). RECORD-ONLY, NOT ENFORCED: nothing on-chain reads this field — the purchase entry never requires it, and an order supplying none of the declared fields is still created. FLOW: the order holder sends the information to the Service's Contact (um) via end-to-end encrypted Messenger, then passes the Contact object ID or the generated WTS Proof here. To ENFORCE delivery on-chain, mount the pattern.required_info_evidence_gate Guard on the forward that hands over goods or funds, asserting order.required_info is some (presence-only — the recorded address value is never a guard condition; accepting the delivered information is the merchant's confirming forward).",
+                  "description": "Contact object ID or WTS Proof object, RECORDED on the Order as evidence that the order holder delivered the private information the Service declares in customer_required (e.g. phone/email/shipping_address). RECORD-ONLY, NOT ENFORCED: nothing on-chain reads this field — the purchase entry never requires it, and an order supplying none of the declared fields is still created. FLOW: the order holder sends the information to the Service's Contact (um) via end-to-end encrypted Messenger, then passes the Contact object ID or the generated WTS Proof here. To ENFORCE delivery on-chain, mount the pattern.required_info_evidence_gate Guard on the forward that hands over goods or funds, asserting order.required_info is some (presence-only — the recorded address value is never a guard condition; accepting the delivered information is the merchant's confirming forward). ATTESTATION-ONLY: the recorded value is self-attested, has no on-chain readback, and MUST NOT be used as a gate condition. PER-ORDER BINDING: when the listing genuinely promises a per-order credential, use the dedicated controlled bind operation (order_required_info_bind) with a Proof whose about is THIS order and whose creator is the order holder — the chain verifies both at write time and rejects a mismatch without writing anything. The reuse model (one Contact Proof across orders) stays the default for this plain record field; pick per listing promise, not per habit.",
                   "type": "string"
                 },
                 "transfer": {
@@ -743,7 +742,7 @@
               ]
             },
             "rewards": {
-              "description": "Service Reward object list. May share the Service's Permission (no conflict check — intentional: rewards are the provider's own incentive tool; only Arbitration requires a different Permission for neutrality).",
+              "description": "Service Reward object list. May share the Service's Permission (no conflict check — intentional: rewards are the provider's own incentive tool; only Arbitration requires a different Permission for neutrality). ⚠️ OP SEMANTICS (SDK translation, etm-mn-261013 E-06): op:'set' executes as clear + add — and reward clear is TIME-LOCKED after publish (E_LOCKED_TIME_NOT_ENOUGH = abort 24: requires pause + setting_lock_duration elapsed, default 30 days). BEFORE publish op:'set' is fine; AFTER publish attach rewards with op:'add' ONLY (add has no lock).",
               "anyOf": [
                 {
                   "type": "object",
@@ -813,7 +812,7 @@
               ]
             },
             "arbitrations": {
-              "description": "Service Arbitration object list. FORMAT: same as repositories — use the `objects` field name inside the operation data. Example: {arbitrations: {op: 'set', objects: [{name: 'my_arb_1'}, {name: 'my_arb_2'}]}}. NOTE: op:'add' is disabled on this surface (its success response can arrive unparseable and reads like a failure); ALWAYS pass the FULL list with op:'set'. Each item is a NameOrAddress (object ID or local mark name). ⚠️ PERMISSION RULE (contract-enforced): every Arbitration bound here MUST use a Permission object DIFFERENT from THIS Service's permission — binding an Arbitration that shares the Service's Permission aborts with E_ARBITRATION_PERMISSION_CONFLICT (error 33). Create a dedicated Permission for each Arbitration BEFORE binding. DESIGN RATIONALE (confirmed intentional): dispute resolution requires NEUTRALITY — an Arbitration controlled by the same Permission admins as the Service it judges would be a conflict of interest. ONLY Arbitration has this check: Reward and Treasury objects MAY share the Service's Permission (they are the provider's own tools).",
+              "description": "Service Arbitration object list. FORMAT: same as repositories — use the `objects` field name inside the operation data. Example: {arbitrations: {op: 'add', objects: [{name: 'my_arb_1'}]}}. Each item is a NameOrAddress (object ID or local mark name). ⚠️ OP SEMANTICS (SDK translation, etm-mn-261013 E-06 + crs P3-1): op:'add' appends and is the ONLY  unlocked path after publish (add has no publish assert, service.move). op:'set' executes as clear + add and REPLACES the entire allowlist (pass EVERY Arbitration that must remain bound) — clear is TIME-LOCKED after publish (E_LOCKED_TIME_NOT_ENOUGH = abort 24: requires pause + setting_lock_duration elapsed, default 30 days), so use op:'set' BEFORE publish only. (The 2026-10-08 op:'add' disable was lifted: the unparseable-success-response SDK bug it responded to is fixed — safeTextJson + big-int conversion.) ⚠️ PERMISSION RULE (contract-enforced): every Arbitration bound here MUST use a Permission object DIFFERENT from THIS Service's permission — binding an Arbitration that shares the Service's Permission aborts with E_ARBITRATION_PERMISSION_CONFLICT (error 33). Create a dedicated Permission for each Arbitration BEFORE binding. DESIGN RATIONALE (confirmed intentional): dispute resolution requires NEUTRALITY — an Arbitration controlled by the same Permission admins as the Service it judges would be a conflict of interest. ONLY Arbitration has this check: Reward and Treasury objects MAY share the Service's Permission (they are the provider's own tools).",
               "anyOf": [
                 {
                   "type": "object",
@@ -925,7 +924,7 @@
                   "description": "Discount magnitude, meaning depends on discount_type. RATES (0): per-10000 off rate — 10 = 0.1% off, 1000 = 10% off, 10000 = 100% off; values above 10000 are rejected. FIXED (1): fixed deduction from the order total in the Service token's smallest units (e.g. 1000000000 = 1 WOW at 9 decimals)."
                 },
                 "benchmark": {
-                  "description": "Minimum order total (in the Service token's smallest units) for the discount to apply — when set, the discount is applied only if the PRE-discount order total is GREATER THAN OR EQUAL TO this value. ✅ OMITTING this field (or passing null, or 0) means NO threshold — the discount applies UNCONDITIONALLY, and that is the common, correct case. This is the on-chain behaviour: benchmark_met = if (option::is_some(&discount.benchmark)) { total >= benchmark } else { true } (wow-framework sources/w/service.move:274-278, inside the discount-application step). ⚠️ SHARP EDGE when you DO set a threshold: a below-threshold order is NOT rejected — it completes at FULL price and the one-time coupon is still consumed and destroyed, so the buyer pays full price AND loses the coupon. Keep thresholds low enough that your typical basket clears them. If the order total is below the discounted price the payment aborts — pay the list price; the excess is returned as a new Coin.",
+                  "description": "Minimum order total (in the Service token's smallest units) for the discount to apply — when set, the discount is applied only if the PRE-discount order total is GREATER THAN OR EQUAL TO this value. ✅ OMITTING this field (or passing null, or 0) means NO threshold — the discount applies UNCONDITIONALLY, and that is the common, correct case. This is the on-chain behaviour: benchmark_met = if (option::is_some(&discount.benchmark)) { total >= benchmark } else { true } (wow-framework sources/w/service.move:274-278, inside the discount-application step). ⚠️ SHARP EDGE when you DO set a threshold: an order below the threshold CANNOT use the coupon — attaching it at buy aborts (error 38: the transaction rolls back, nothing is paid, and the coupon is preserved); dropping it means the buyer pays FULL price and keeps the coupon for a future order. Keep thresholds low enough that your typical basket clears them. If the order total is below the discounted price the payment aborts — pay the list price; the excess is returned as a new Coin.",
                   "anyOf": [
                     {
                       "type": "number"
@@ -936,9 +935,13 @@
                   ]
                 },
                 "time_ms_start": {
-                  "description": "Discount start time (milliseconds)",
+                  "description": "Discount start time (milliseconds). ⚠️ A FUTURE start mints a coupon that is listed as usable but CANNOT be spent until the window opens (every spend aborts E_DISCOUNT_TIME_INVALID / buy code 35). Legitimate for scheduled campaigns — the mint carries a loud warning; do not distribute the coupon to recipients before the window opens.",
                   "type": "number"
                 },
+                "allow_zero_total": {
+                  "description": "FREE-TIER ESCAPE HATCH (BKN-P2): a coupon that would drive the final total to 0 (MINUS value >= the cheapest SKU price, or RATES value 10000) is BLOCKED at issuance by default. Set true ONLY to declare a deliberate zero-price product: the issuance then proceeds with a disclosure that zero-total orders carry an empty escrow and NEVER trigger reward/incentive rails. Omit (default false) for normal discounts.",
+                  "type": "boolean"
+                },
                 "time_ms_end": {
                   "type": "number",
                   "description": "Discount end time (milliseconds)"
@@ -979,7 +982,7 @@
                     "check_all_founded"
                   ],
                   "additionalProperties": false,
-                  "description": "Discount recipient"
+                  "description": "Discount recipient. OBJECT FORM REQUIRED (T9 az3-mn-261013): a bare string is REJECTED here — pass {entities: [{name_or_address: '<account_name_or_0x_address>'}]} (example: {entities: [{name_or_address: 'buyer1'}]}). Total coupons delivered = count × number of recipients."
                 },
                 "transferable": {
                   "type": "boolean",
@@ -1016,7 +1019,7 @@
                     "type": "string",
                     "description": "Account/Object name or ID. If specifying an account, use empty string '' for the default account. If it starts with '0x', it will be treated as an ID. Otherwise, it will be treated as a name (max 64 bcs characters)."
                   },
-                  "description": "Coupon object IDs or local names to transfer. You MUST own these Discount objects. Each coupon must have been minted with transferable:true — otherwise the chain aborts with E_DISCOUNT_NOT_TRANSFERABLE (semantic code service:10). Coupons stay single-use; the transfer does not change that. Max 1000 coupons per call (SDK limit MAX_DISCOUNT_COUNT_ONCE)."
+                  "description": "Coupon object IDs or local names to transfer. You MUST own these Discount objects. Each coupon must have been minted with transferable:true — otherwise the chain aborts with E_DISCOUNT_NOT_TRANSFERABLE (semantic code service:10). Coupons stay single-use; the transfer does not change that. There is NO batch entry on-chain — each coupon is delivered by its own singleton call (fanned out automatically); prefer ONE coupon per transaction for coupon migrations (SDK cap MAX_DISCOUNT_COUNT_ONCE still applies)."
                 },
                 "recipient": {
                   "type": "object",
@@ -1427,7 +1430,7 @@
               ]
             },
             "compensation_fund_add": {
-              "description": "Deposit funds into the Service compensation_fund. Used to pay indemnity when arbitration resolves in customer's favor. FORMAT: {balance: <amount_in_smallest_unit>} — the field name is 'balance' (NOT 'amount'). The token type and precision are determined by the Service object's type_parameter (the generic type set when the Service was created). For WOW (9 decimals): {balance: 1000000000} = 1 WOW. For tokens with different decimals, adjust accordingly (e.g. USDC has 6 decimals, so {balance: 1000000} = 1 USDC). REQUIRES: Permission index 315 (SERVICE_COMPENSATION_FUND_DEPOSIT) must be granted to the calling account first. COMMON MISTAKE: using {amount: ...} or {amount: ..., type: 'WOW'} — these will fail. The correct field is 'balance'.",
+              "description": "Deposit funds into the Service compensation_fund. Used to pay indemnity when arbitration resolves in customer's favor. FORMAT: {balance: <amount_in_smallest_unit>} — the field name is 'balance' (NOT 'amount'). The token type and precision are determined by the Service object's type_parameter (the generic type set when the Service was created). For WOW (9 decimals): {balance: 1000000000} = 1 WOW. For tokens with different decimals, adjust accordingly (e.g. USDC has 6 decimals, so {balance: 1000000} = 1 USDC). REQUIRES: Permission index 315 (SERVICE_COMPENSATION_FUND_DEPOSIT) for NON-ADMIN callers. The Permission owner and admin short-circuit every builtin index check (no grant needed); only non-admin callers must be granted 315 first. COMMON MISTAKE: using {amount: ...} or {amount: ..., type: 'WOW'} — these will fail. The correct field is 'balance'.",
               "anyOf": [
                 {
                   "type": "object",
@@ -1630,6 +1633,7 @@
               "description": "Unwrap CoinWrapper objects and other objects received by this Service object and send them to the owner of its Permission object.\n\nACCEPTED FORMATS (F-06 unified receive operation block):\n• 'recently' (string literal) — auto-query and receive ALL recently received objects.\n  Use this for the common case: \"withdraw everything the object has received.\"\n  Example: receive: 'recently'\n• ReceivedNormal[] (array) — explicit list of received objects to unwrap.\n  Use this when you want to receive specific objects only (not all).\n  Example: receive: [{id: '0xobj1', type: '0x2::coin::Coin<0x2::wow::WOW>'}]\n• ReceivedBalance ({token_type, balance, received: [{id, balance, payment}]}) —\n  receive a balance record from a specific Payment/payer.\n  Use this for precise balance targeting (advanced — usually after querying\n  the object's received history via query_received).\n  Example: receive: {token_type: '0x2::wow::WOW', balance: 1000000, received: [{id: '0xrec1', balance: 1000000, payment: '0xpay1'}]}\nANTI-PATTERN: do NOT wrap in {result: ...} — pass the value directly.",
               "anyOf": [
                 {
+                  "minItems": 1,
                   "type": "array",
                   "items": {
                     "type": "object",
@@ -1786,7 +1790,7 @@
               "type": "boolean"
             },
             "publish": {
-              "description": "Whether to publish the Service. After publishing, customers can place orders. ⚠️ PRE-PUBLISH EVALUATION (mandatory, stop-node fund-exit coverage): service.allocators MUST cover EVERY stop node of the bound Machine — every node where an order can rest (terminal nodes AND waiting states: review windows, handovers, open disputes) needs BOTH a withdrawal (release) AND a refund allocation rule, and each mapping guard must be satisfiable in BOTH directions. An uncovered stop node strands escrowed funds permanently (order_allocators is L1-immutable). Review-gated release is a TIME VARIANT inside ONE exit guard ('reviewed OR unreviewed after T'), never parallel outgoing edges (race deadlock). See pattern.fund_flow_stop_node_coverage + rules R-M5-03 / R-M5-09. VERIFIED against Move source + SDK (4-level immutability matrix):\n  L1 — PERMANENTLY LOCKED after publish (assert!(!bPublished), no pause+lock exception):\n    • machine          (workflow template)\n    • order_allocators (fund distribution rules — full replacement locked; APPEND-ONLY additions remain possible via order_allocators_append, prefix/threshold frozen)\n  L2 — TIME-LOCKED after publish (unpublished-state check — requires pause + setting_lock_duration elapsed):\n    • arbitrations remove/clear (dispute resolution objects)\n    • rewards remove/clear       (reward objects)\n    • compensation_fund_withdraw (withdraw ALL funds)\n  L3 — REMAIN MUTABLE after publish (no SDK check, no Move check):\n    • arbitrations add, rewards add (no assert — can add after publish)\n    • buy_guard, sales, discount, description, location, pause, repositories,\n    • compensation_fund_add, setting_lock_duration_add, customer_required, um (Contact)\nThese 2 L1-LOCKED fields (machine/order_allocators) MUST be set BEFORE publish=true.\narbitrations/rewards can be ADDED after publish but remove/clear requires pause+lock.\n\n⚠️ COMPENSATION_FUND + ARBITRATION LINKAGE:\n  At publish time, if compensation_fund > 0, Arbitration MUST be bound:\n    if (balance::value(&self.compensation_fund) > 0) {\n        assert!(arbitration_count > 0, E_ARBITRATION_NOT_SET_WITH_COMPENSATION_FUND);\n    }\n  The MCP handler enforces this as a HARD PRE-CHECK: if publish=true AND compensation_fund_add   is set in the same call AND arbitrations is empty, the call is REJECTED before submission.   If compensation_fund was deposited in a prior call, a SOFT WARNING is issued.\n\nSCHEMA-03 / P0-01 fix — DEPLOYMENT WORKFLOW (two-phase, avoids circular dependency):\n  Phase 1 — CREATE (no publish): object={name:'my-service', type_parameter, permission} + machine + order_allocators + arbitrations.\n    NOTE: buy_guard can use a LocalMark NAME (not address) to break the Guard→Service circular dependency.\n    The name is resolved to an address at transaction build time.\n  Phase 2 — PUBLISH: object='my-service' (string ref) + publish=true.\n    All L1-LOCKED fields must be set in Phase 1; Phase 2 only flips the publish flag.\n  Post-publish updates: buy_guard, sales, description, repositories (add), rewards (add), arbitrations (add), etc.",
+              "description": "Whether to publish the Service. After publishing, customers can place orders. ⚠️ PRE-PUBLISH EVALUATION (mandatory, stop-node fund-exit coverage): service.allocators MUST cover EVERY stop node of the bound Machine — every node where an order can rest (terminal nodes AND waiting states: review windows, handovers, open disputes) needs BOTH a withdrawal (release) AND a refund allocation rule, and each mapping guard must be satisfiable in BOTH directions. An uncovered stop node strands escrowed funds permanently (order_allocators is L1-immutable). Review-gated release is a TIME VARIANT inside ONE exit guard ('reviewed OR unreviewed after T'), never parallel outgoing edges (race deadlock). See pattern.fund_flow_stop_node_coverage + rules R-M5-03 / R-M5-09. VERIFIED against Move source + SDK (4-level immutability matrix):\n  L1 — PERMANENTLY LOCKED after publish (assert!(!bPublished), no pause+lock exception):\n    • machine          (workflow template)\n    • order_allocators (fund distribution rules — full replacement locked; APPEND-ONLY additions remain possible via order_allocators_append, prefix/threshold frozen)\n  L2 — TIME-LOCKED after publish (unpublished-state check — requires pause + setting_lock_duration elapsed):\n    • arbitrations remove/clear (dispute resolution objects)\n    • rewards remove/clear       (reward objects)\n    • compensation_fund_withdraw (withdraw ALL funds)\n    RATIONALE (deliberate design, NOT a defect): these three REVOKE buyer-facing guarantees that placed orders already rely on — the lock protects buyer expectations. Daily operations (sales add/remove, price, stock, description) are NEVER gated by this lock.\n  L3 — REMAIN MUTABLE after publish (no SDK check, no Move check):\n    • arbitrations add, rewards add (no assert — can add after publish)\n    • buy_guard, sales, discount, description, location, pause, repositories,\n    • compensation_fund_add, setting_lock_duration_add, customer_required, um (Contact)\nThese 2 L1-LOCKED fields (machine/order_allocators) MUST be set BEFORE publish=true.\narbitrations/rewards can be ADDED after publish but remove/clear requires pause+lock.\n\n⚠️ COMPENSATION_FUND + ARBITRATION LINKAGE:\n  At publish time, if compensation_fund > 0, Arbitration MUST be bound:\n    if (balance::value(&self.compensation_fund) > 0) {\n        assert!(arbitration_count > 0, E_ARBITRATION_NOT_SET_WITH_COMPENSATION_FUND);\n    }\n  The MCP handler enforces this as a HARD PRE-CHECK: if publish=true AND compensation_fund_add   is set in the same call AND arbitrations is empty, the call is REJECTED before submission.   If compensation_fund was deposited in a prior call, a SOFT WARNING is issued.\n\nSCHEMA-03 / P0-01 fix — DEPLOYMENT WORKFLOW (two-phase, avoids circular dependency):\n  Phase 1 — CREATE (no publish): object={name:'my-service', type_parameter, permission} + machine + order_allocators + arbitrations.\n    NOTE: buy_guard can use a LocalMark NAME (not address) to break the Guard→Service circular dependency.\n    The name is resolved to an address at transaction build time.\n  Phase 2 — PUBLISH: object='my-service' (string ref) + publish=true.\n    All L1-LOCKED fields must be set in Phase 1; Phase 2 only flips the publish flag.\n  Post-publish updates: buy_guard, sales, description, repositories (add), rewards (add), arbitrations (add), etc.",
               "type": "boolean"
             }
           },
@@ -1941,6 +1945,7 @@
                           "description": "Whether this table item's value is submitted dynamically at Guard trigger time (alloc_by_guard call). \n\ntrue = value is submitted by the caller when triggering the Guard. Use for runtime-context-dependent values like order address, user address. The 'value' field is ignored when b_submission=true; the caller must provide it via submissions[]. \n\nfalse = value is static, set at Guard creation time. Use for values known when the Guard is created: expected node names, expected merchant address, expected service address. The 'value' field must be populated and will be stored on-chain permanently. \n\nRule of thumb: if the value is the SAME for all future Guard triggers, use false. If the value DIFFERS per trigger (e.g., which order to release funds for), use true."
                         },
                         "value_type": {
+                          "description": "Type of the value stored in `value`. One of: Bool(0), Address(1), String(2), U8(3), U16(4), U32(5), U64(6), U128(7), U256(8), VecBool(9), VecAddress(10), VecString(11), VecU8(12), VecU16(13), VecU32(14), VecU64(15), VecU128(16), VecU256(17), VecVecU8(18). When value_type=Address (1), the `value` field accepts a hex address string, a LocalMark name, or an AccountOrMark_Address object — see `value` field description for details.",
                           "anyOf": [
                             {
                               "type": "number",
@@ -2227,11 +2232,10 @@
                               "const": "vecvecu8",
                               "description": "vecvecu8"
                             }
-                          ],
-                          "description": "Type of the value stored in `value`. One of: Bool(0), Address(1), String(2), U8(3), U16(4), U32(5), U64(6), U128(7), U256(8), VecBool(9), VecAddress(10), VecString(11), VecU8(12), VecU16(13), VecU32(14), VecU64(15), VecU128(16), VecU256(17), VecVecU8(18). When value_type=Address (1), the `value` field accepts a hex address string, a LocalMark name, or an AccountOrMark_Address object — see `value` field description for details."
+                          ]
                         },
                         "value": {
-                          "description": "The actual value data. Format depends on `value_type`:\n• Bool: true/false (boolean)\n• Address (CRITICAL — string 'Address' is NOT a placeholder): a hex address (e.g. '0x1234...'), a LocalMark name (e.g. 'my-service' — resolved to address at evaluation time), an AccountOrMark_Address object (e.g. {name_or_address:'my-service'}), or system shorthand ('0xaaa' = EntityLinker, '0xaab' = EntityRegistrar). The literal string 'Address' itself is INVALID — it would be treated as a non-existent LocalMark name and fail. Example: value='0x2::wow::WOW<address>' or value='my-permission'.\n• String: any string\n• U8/U16/U32/U64/U128/U256: number or numeric string (e.g. 42 or '42')\n• Vec* types: arrays of the corresponding element type\nREQUIRED when b_submission=false. OPTIONAL when b_submission=true (value is supplied at evaluation time by user submission).",
+                          "description": "The actual value data. Format depends on `value_type`:\nTYPE-CLASS SPLIT (F03): scalar value_types (Bool/Address/String/U8..U256) take ONE scalar value (boolean/address string/name/string/number); Vec* value_types (VecBool/VecAddress/VecString/VecU8..VecU256) take a REAL JSON ARRAY of elements — a string (even a Python-repr-looking \"[a, b]\") fails serialization with 'value_type=10 (array) requires a array value'. The historical 'static values must be strings' rule applied to SCALARS only.\n• Bool: true/false (boolean)\n• Address (CRITICAL — string 'Address' is NOT a placeholder): a hex address (e.g. '0x1234...'), a LocalMark name (e.g. 'my-service' — resolved to address at evaluation time), an AccountOrMark_Address object (e.g. {name_or_address:'my-service'}), or system shorthand ('0xaaa' = EntityLinker, '0xaab' = EntityRegistrar). The literal string 'Address' itself is INVALID — it would be treated as a non-existent LocalMark name and fail. Example: value='0x2::wow::WOW<address>' or value='my-permission'.\n• String: any string\n• U8/U16/U32/U64/U128/U256: number or numeric string (e.g. 42 or '42')\n• Vec* types: arrays of the corresponding element type\nPROTOCOL SLOT 254 (once, b_submission Address): the submitted value MUST be the ORDER address bound to the operation (payment_info.for_object) — NEVER the Progress/Machine/Arb object the call is advancing. The chain asserts submission(254) == the order address; a Progress address fails the task binding and the forward aborts.\nREQUIRED when b_submission=false. OPTIONAL when b_submission=true (value is supplied at evaluation time by user submission).",
                           "anyOf": [
                             {
                               "type": "boolean"
@@ -2390,7 +2394,7 @@
                       "additionalProperties": false,
                       "description": "Guard table item (QUERY/OUTPUT form — includes auto-derived object_type field)"
                     },
-                    "description": "User-submitted data matching the Guard's required fields. Relation: structure must match the Guard table's column definitions. Example: [{field:'delivery_proof', value:'Qm...'}]"
+                    "description": "User-submitted data matching the Guard's required fields. Relation: structure must match the Guard table's column definitions. Example: [{field:'delivery_proof', value:'Qm...'}] PROTOCOL SLOT 254: when the Guard table has identifier 254, submit the ORDER address bound to the operation (payment_info.for_object) — never the Progress/Machine object being advanced."
                   }
                 },
                 "required": [
@@ -2400,7 +2404,7 @@
                 "additionalProperties": false,
                 "description": "One Guard's submission data: the Guard to verify plus the user-provided data that satisfies its requirements."
               },
-              "description": "User-submitted data for each Guard. Relation: one entry per Guard in the guard array; fill submission fields and resubmit via call_with_submission. PLACEMENT: this `submission` field is at the SAME level as `data` and `env` in the operation input — NOT inside `data.data`. Example structure: {tool:'onchain_operations', data:{operation_type:'order', data:{object:'my_order', progress:{...}}}, submission:{type:'submission', guard:[...], submission:[...]}}"
+              "description": "User-submitted data for each Guard. Relation: one entry per Guard in the guard array; fill submission fields and resubmit via call_with_submission. SLOT COMPLETENESS (T8 az3-mn-261013): the submission array must cover EVERY b_submission slot in the Guard's table, not just one — a Guard whose table has multiple b_submission entries (e.g. slot 0 + slot 254) aborts on-chain with 'Submission ID not found' (abort 11) when any b_submission slot has no matching entry. Check the Guard's table (each entry with b_submission needs its identifier filled) and include ALL of them. PLACEMENT: this `submission` field is at the SAME level as `data` and `env` in the operation input — NOT inside `data.data`. Example structure: {tool:'onchain_operations', data:{operation_type:'order', data:{object:'my_order', progress:{...}}}, submission:{type:'submission', guard:[...], submission:[...]}}"
             }
           },
           "required": [
@@ -2780,7 +2784,7 @@
                                 "description": "Node name. ⚠️ T1 LOSSY POINT (B-1): do NOT name a node 'refund'/'refunded'/'deposit_refunded'/'deposit_deducted'/'disputed'/'cancelled' to model fund return or dispute — funds MUST move through the Allocation waterfall (order_allocators), NOT through Machine terminal nodes (R-M1-11). A node named 'refund' is a semantic lie: reaching it does not move funds by itself. Model refund as a refund-to-Order Allocator slot (a { Signer } or Order-owner recipient), and model dispute as the bound Arbitration, not a Machine node."
                               },
                               "role": {
-                                "description": "OPTIONAL business role tag for display layers and template reuse (E-9). SESSION METADATA ONLY — the on-chain node struct carries name+pairs, so this is NOT stored on-chain; the MCP echoes it back as a NODE ROLES constraint warning and seed templates annotate it. Semantics: draft = pre-start cancellable, active = work in progress, delivered = work submitted awaiting acceptance, final = settlement endpoint (advancing here should trigger the Allocation fund release), cancelled = refund endpoint (advancing here should trigger the refund Allocator).",
+                                "description": "OPTIONAL business role tag for display layers and template reuse (E-9). SESSION METADATA ONLY — the on-chain node struct carries name+pairs, so this is NOT stored on-chain; the MCP echoes it back as a NODE ROLES constraint warning and seed templates annotate it. Semantics: draft = pre-start cancellable, active = work in progress, delivered = work submitted awaiting acceptance, final = settlement endpoint, cancelled = refund endpoint. ⚠️ Advancing to ANY node never moves funds by itself: settlement/refund requires an explicit allocation call (alloc by guard) AFTER the terminal node is reached. Nothing auto-settles on-chain — plan the settlement trigger (merchant, keeper, or buyer action) explicitly in the template.",
                                 "type": "string",
                                 "enum": [
                                   "draft",
@@ -2967,7 +2971,7 @@
                                 "description": "Node name. ⚠️ T1 LOSSY POINT (B-1): do NOT name a node 'refund'/'refunded'/'deposit_refunded'/'deposit_deducted'/'disputed'/'cancelled' to model fund return or dispute — funds MUST move through the Allocation waterfall (order_allocators), NOT through Machine terminal nodes (R-M1-11). A node named 'refund' is a semantic lie: reaching it does not move funds by itself. Model refund as a refund-to-Order Allocator slot (a { Signer } or Order-owner recipient), and model dispute as the bound Arbitration, not a Machine node."
                               },
                               "role": {
-                                "description": "OPTIONAL business role tag for display layers and template reuse (E-9). SESSION METADATA ONLY — the on-chain node struct carries name+pairs, so this is NOT stored on-chain; the MCP echoes it back as a NODE ROLES constraint warning and seed templates annotate it. Semantics: draft = pre-start cancellable, active = work in progress, delivered = work submitted awaiting acceptance, final = settlement endpoint (advancing here should trigger the Allocation fund release), cancelled = refund endpoint (advancing here should trigger the refund Allocator).",
+                                "description": "OPTIONAL business role tag for display layers and template reuse (E-9). SESSION METADATA ONLY — the on-chain node struct carries name+pairs, so this is NOT stored on-chain; the MCP echoes it back as a NODE ROLES constraint warning and seed templates annotate it. Semantics: draft = pre-start cancellable, active = work in progress, delivered = work submitted awaiting acceptance, final = settlement endpoint, cancelled = refund endpoint. ⚠️ Advancing to ANY node never moves funds by itself: settlement/refund requires an explicit allocation call (alloc by guard) AFTER the terminal node is reached. Nothing auto-settles on-chain — plan the settlement trigger (merchant, keeper, or buyer action) explicitly in the template.",
                                 "type": "string",
                                 "enum": [
                                   "draft",
@@ -3481,7 +3485,7 @@
               ]
             },
             "pause": {
-              "description": "Whether to pause generating new Progress objects.",
+              "description": "Whether to pause generating new Progress objects. Chain truth (machine.move:657, the progress-generation entry asserts !bPaused): the pause gates NEW order entry only — in-flight Progress keeps advancing and settling (yfd-ln F-13 misread this as a freeze switch; it never was one).",
               "type": "boolean"
             },
             "publish": {
@@ -3492,6 +3496,7 @@
               "description": "Unwrap CoinWrapper objects and other objects received by this Machine object and send them to the owner of its Permission object.\n\nACCEPTED FORMATS (F-06 unified receive operation block):\n• 'recently' (string literal) — auto-query and receive ALL recently received objects.\n  Use this for the common case: \"withdraw everything the object has received.\"\n  Example: receive: 'recently'\n• ReceivedNormal[] (array) — explicit list of received objects to unwrap.\n  Use this when you want to receive specific objects only (not all).\n  Example: receive: [{id: '0xobj1', type: '0x2::coin::Coin<0x2::wow::WOW>'}]\n• ReceivedBalance ({token_type, balance, received: [{id, balance, payment}]}) —\n  receive a balance record from a specific Payment/payer.\n  Use this for precise balance targeting (advanced — usually after querying\n  the object's received history via query_received).\n  Example: receive: {token_type: '0x2::wow::WOW', balance: 1000000, received: [{id: '0xrec1', balance: 1000000, payment: '0xpay1'}]}\nANTI-PATTERN: do NOT wrap in {result: ...} — pass the value directly.",
               "anyOf": [
                 {
+                  "minItems": 1,
                   "type": "array",
                   "items": {
                     "type": "object",
@@ -3770,6 +3775,7 @@
                           "description": "Whether this table item's value is submitted dynamically at Guard trigger time (alloc_by_guard call). \n\ntrue = value is submitted by the caller when triggering the Guard. Use for runtime-context-dependent values like order address, user address. The 'value' field is ignored when b_submission=true; the caller must provide it via submissions[]. \n\nfalse = value is static, set at Guard creation time. Use for values known when the Guard is created: expected node names, expected merchant address, expected service address. The 'value' field must be populated and will be stored on-chain permanently. \n\nRule of thumb: if the value is the SAME for all future Guard triggers, use false. If the value DIFFERS per trigger (e.g., which order to release funds for), use true."
                         },
                         "value_type": {
+                          "description": "Type of the value stored in `value`. One of: Bool(0), Address(1), String(2), U8(3), U16(4), U32(5), U64(6), U128(7), U256(8), VecBool(9), VecAddress(10), VecString(11), VecU8(12), VecU16(13), VecU32(14), VecU64(15), VecU128(16), VecU256(17), VecVecU8(18). When value_type=Address (1), the `value` field accepts a hex address string, a LocalMark name, or an AccountOrMark_Address object — see `value` field description for details.",
                           "anyOf": [
                             {
                               "type": "number",
@@ -4056,11 +4062,10 @@
                               "const": "vecvecu8",
                               "description": "vecvecu8"
                             }
-                          ],
-                          "description": "Type of the value stored in `value`. One of: Bool(0), Address(1), String(2), U8(3), U16(4), U32(5), U64(6), U128(7), U256(8), VecBool(9), VecAddress(10), VecString(11), VecU8(12), VecU16(13), VecU32(14), VecU64(15), VecU128(16), VecU256(17), VecVecU8(18). When value_type=Address (1), the `value` field accepts a hex address string, a LocalMark name, or an AccountOrMark_Address object — see `value` field description for details."
+                          ]
                         },
                         "value": {
-                          "description": "The actual value data. Format depends on `value_type`:\n• Bool: true/false (boolean)\n• Address (CRITICAL — string 'Address' is NOT a placeholder): a hex address (e.g. '0x1234...'), a LocalMark name (e.g. 'my-service' — resolved to address at evaluation time), an AccountOrMark_Address object (e.g. {name_or_address:'my-service'}), or system shorthand ('0xaaa' = EntityLinker, '0xaab' = EntityRegistrar). The literal string 'Address' itself is INVALID — it would be treated as a non-existent LocalMark name and fail. Example: value='0x2::wow::WOW<address>' or value='my-permission'.\n• String: any string\n• U8/U16/U32/U64/U128/U256: number or numeric string (e.g. 42 or '42')\n• Vec* types: arrays of the corresponding element type\nREQUIRED when b_submission=false. OPTIONAL when b_submission=true (value is supplied at evaluation time by user submission).",
+                          "description": "The actual value data. Format depends on `value_type`:\nTYPE-CLASS SPLIT (F03): scalar value_types (Bool/Address/String/U8..U256) take ONE scalar value (boolean/address string/name/string/number); Vec* value_types (VecBool/VecAddress/VecString/VecU8..VecU256) take a REAL JSON ARRAY of elements — a string (even a Python-repr-looking \"[a, b]\") fails serialization with 'value_type=10 (array) requires a array value'. The historical 'static values must be strings' rule applied to SCALARS only.\n• Bool: true/false (boolean)\n• Address (CRITICAL — string 'Address' is NOT a placeholder): a hex address (e.g. '0x1234...'), a LocalMark name (e.g. 'my-service' — resolved to address at evaluation time), an AccountOrMark_Address object (e.g. {name_or_address:'my-service'}), or system shorthand ('0xaaa' = EntityLinker, '0xaab' = EntityRegistrar). The literal string 'Address' itself is INVALID — it would be treated as a non-existent LocalMark name and fail. Example: value='0x2::wow::WOW<address>' or value='my-permission'.\n• String: any string\n• U8/U16/U32/U64/U128/U256: number or numeric string (e.g. 42 or '42')\n• Vec* types: arrays of the corresponding element type\nPROTOCOL SLOT 254 (once, b_submission Address): the submitted value MUST be the ORDER address bound to the operation (payment_info.for_object) — NEVER the Progress/Machine/Arb object the call is advancing. The chain asserts submission(254) == the order address; a Progress address fails the task binding and the forward aborts.\nREQUIRED when b_submission=false. OPTIONAL when b_submission=true (value is supplied at evaluation time by user submission).",
                           "anyOf": [
                             {
                               "type": "boolean"
@@ -4219,7 +4224,7 @@
                       "additionalProperties": false,
                       "description": "Guard table item (QUERY/OUTPUT form — includes auto-derived object_type field)"
                     },
-                    "description": "User-submitted data matching the Guard's required fields. Relation: structure must match the Guard table's column definitions. Example: [{field:'delivery_proof', value:'Qm...'}]"
+                    "description": "User-submitted data matching the Guard's required fields. Relation: structure must match the Guard table's column definitions. Example: [{field:'delivery_proof', value:'Qm...'}] PROTOCOL SLOT 254: when the Guard table has identifier 254, submit the ORDER address bound to the operation (payment_info.for_object) — never the Progress/Machine object being advanced."
                   }
                 },
                 "required": [
@@ -4229,7 +4234,7 @@
                 "additionalProperties": false,
                 "description": "One Guard's submission data: the Guard to verify plus the user-provided data that satisfies its requirements."
               },
-              "description": "User-submitted data for each Guard. Relation: one entry per Guard in the guard array; fill submission fields and resubmit via call_with_submission. PLACEMENT: this `submission` field is at the SAME level as `data` and `env` in the operation input — NOT inside `data.data`. Example structure: {tool:'onchain_operations', data:{operation_type:'order', data:{object:'my_order', progress:{...}}}, submission:{type:'submission', guard:[...], submission:[...]}}"
+              "description": "User-submitted data for each Guard. Relation: one entry per Guard in the guard array; fill submission fields and resubmit via call_with_submission. SLOT COMPLETENESS (T8 az3-mn-261013): the submission array must cover EVERY b_submission slot in the Guard's table, not just one — a Guard whose table has multiple b_submission entries (e.g. slot 0 + slot 254) aborts on-chain with 'Submission ID not found' (abort 11) when any b_submission slot has no matching entry. Check the Guard's table (each entry with b_submission needs its identifier filled) and include ALL of them. PLACEMENT: this `submission` field is at the SAME level as `data` and `env` in the operation input — NOT inside `data.data`. Example structure: {tool:'onchain_operations', data:{operation_type:'order', data:{object:'my_order', progress:{...}}}, submission:{type:'submission', guard:[...], submission:[...]}}"
             }
           },
           "required": [
@@ -4437,7 +4442,7 @@
                   ]
                 },
                 "message": {
-                  "description": "Operation result message.",
+                  "description": "Message attached to the forward advance; recorded permanently in on-chain History. REVIEW-TYPE FORWARDS: de-facto REQUIRED here — the review text is this message (the reward claim guard does not read it, but a no-message Review forward has failed repeatedly on mainnet until the message was added). Always include message: '<review text>' when advancing a Review-named forward.",
                   "type": "string"
                 }
               },
@@ -4451,6 +4456,7 @@
               "description": "Unwrap CoinWrapper objects and other objects received by this Progress object and send them to the owner of the bound Machine's Permission object.\n\nACCEPTED FORMATS (F-06 unified receive operation block):\n• 'recently' (string literal) — auto-query and receive ALL recently received objects.\n  Use this for the common case: \"withdraw everything the object has received.\"\n  Example: receive: 'recently'\n• ReceivedNormal[] (array) — explicit list of received objects to unwrap.\n  Use this when you want to receive specific objects only (not all).\n  Example: receive: [{id: '0xobj1', type: '0x2::coin::Coin<0x2::wow::WOW>'}]\n• ReceivedBalance ({token_type, balance, received: [{id, balance, payment}]}) —\n  receive a balance record from a specific Payment/payer.\n  Use this for precise balance targeting (advanced — usually after querying\n  the object's received history via query_received).\n  Example: receive: {token_type: '0x2::wow::WOW', balance: 1000000, received: [{id: '0xrec1', balance: 1000000, payment: '0xpay1'}]}\nANTI-PATTERN: do NOT wrap in {result: ...} — pass the value directly.",
               "anyOf": [
                 {
+                  "minItems": 1,
                   "type": "array",
                   "items": {
                     "type": "object",
@@ -4564,6 +4570,63 @@
                   "const": "recently"
                 }
               ]
+            },
+            "review": {
+              "description": "Structured review composite (REVIEW component): after this forward succeeds, write the structured reputation record into a Repository (+ optional reward claim), reusing this call's passport submission. Requires operate (the chain fires only after a successful forward).",
+              "type": "object",
+              "properties": {
+                "repository": {
+                  "type": "string",
+                  "description": "Repository object ID or name that stores the structured reputation record. The matching policy key must accept a JSON payload {rating, tags?, orderRef?}. id_from options (C30): Signer = one record per BUYER (rewrite = own-record overwrite); None + a write_guard with id_from_submission = one record per bound OBJECT/ORDER (the row id is the raw bound Order address — the verified once-per-order review shape); Clock = append-only log. PREFERRED: the service's review repository EMITTED BY THE BUILD PLAN when review_mode was set (AR-1) — resolve it from the service's repositories list (service_panorama / onchain_objects, name suffix '_review') — never guess an address."
+                },
+                "key": {
+                  "description": "Data key (policy rule name) to write under. Default 'review'.",
+                  "type": "string"
+                },
+                "rating": {
+                  "type": "string",
+                  "enum": [
+                    "like",
+                    "dislike"
+                  ],
+                  "description": "Buyer rating. Ratings NEVER affect settlement (the settlement anchor is the delivered moment, not this review) — this is reputation only."
+                },
+                "tags": {
+                  "description": "Optional structured tags for statistics and market exposure (e.g. ['fast_delivery','good_quality']).",
+                  "type": "array",
+                  "items": {
+                    "type": "string"
+                  }
+                },
+                "reward_claim": {
+                  "description": "Optional reward claim chained after the repository write (fail-isolated: a failed claim never undoes the write or the forward).",
+                  "type": "object",
+                  "properties": {
+                    "reward": {
+                      "type": "string",
+                      "description": "Reward object ID or name to claim from. Resolve it from the service's rewards list (service_panorama / service_quote reward_terms) — never guess an address."
+                    },
+                    "guard": {
+                      "type": "string",
+                      "description": "Guard object that gates this claim. MUST be covered by this call's passport submission (one of the submission.guard entries, same address), otherwise the chain aborts."
+                    }
+                  },
+                  "required": [
+                    "reward",
+                    "guard"
+                  ],
+                  "additionalProperties": false
+                },
+                "write_guard": {
+                  "description": "Policy write guard for the repository write. DEFAULT: the first guard of this call's passport submission (the review-edge credential). Required when the matched Repository policy uses id_from=None with declared guards (the chain aborts with 22, E_POLICY_WRITE_GUARD_REQUIRED, otherwise). The guard MUST be covered by the passport submission.",
+                  "type": "string"
+                }
+              },
+              "required": [
+                "repository",
+                "rating"
+              ],
+              "additionalProperties": false
             }
           },
           "required": [
@@ -4717,6 +4780,7 @@
                           "description": "Whether this table item's value is submitted dynamically at Guard trigger time (alloc_by_guard call). \n\ntrue = value is submitted by the caller when triggering the Guard. Use for runtime-context-dependent values like order address, user address. The 'value' field is ignored when b_submission=true; the caller must provide it via submissions[]. \n\nfalse = value is static, set at Guard creation time. Use for values known when the Guard is created: expected node names, expected merchant address, expected service address. The 'value' field must be populated and will be stored on-chain permanently. \n\nRule of thumb: if the value is the SAME for all future Guard triggers, use false. If the value DIFFERS per trigger (e.g., which order to release funds for), use true."
                         },
                         "value_type": {
+                          "description": "Type of the value stored in `value`. One of: Bool(0), Address(1), String(2), U8(3), U16(4), U32(5), U64(6), U128(7), U256(8), VecBool(9), VecAddress(10), VecString(11), VecU8(12), VecU16(13), VecU32(14), VecU64(15), VecU128(16), VecU256(17), VecVecU8(18). When value_type=Address (1), the `value` field accepts a hex address string, a LocalMark name, or an AccountOrMark_Address object — see `value` field description for details.",
                           "anyOf": [
                             {
                               "type": "number",
@@ -5003,11 +5067,10 @@
                               "const": "vecvecu8",
                               "description": "vecvecu8"
                             }
-                          ],
-                          "description": "Type of the value stored in `value`. One of: Bool(0), Address(1), String(2), U8(3), U16(4), U32(5), U64(6), U128(7), U256(8), VecBool(9), VecAddress(10), VecString(11), VecU8(12), VecU16(13), VecU32(14), VecU64(15), VecU128(16), VecU256(17), VecVecU8(18). When value_type=Address (1), the `value` field accepts a hex address string, a LocalMark name, or an AccountOrMark_Address object — see `value` field description for details."
+                          ]
                         },
                         "value": {
-                          "description": "The actual value data. Format depends on `value_type`:\n• Bool: true/false (boolean)\n• Address (CRITICAL — string 'Address' is NOT a placeholder): a hex address (e.g. '0x1234...'), a LocalMark name (e.g. 'my-service' — resolved to address at evaluation time), an AccountOrMark_Address object (e.g. {name_or_address:'my-service'}), or system shorthand ('0xaaa' = EntityLinker, '0xaab' = EntityRegistrar). The literal string 'Address' itself is INVALID — it would be treated as a non-existent LocalMark name and fail. Example: value='0x2::wow::WOW<address>' or value='my-permission'.\n• String: any string\n• U8/U16/U32/U64/U128/U256: number or numeric string (e.g. 42 or '42')\n• Vec* types: arrays of the corresponding element type\nREQUIRED when b_submission=false. OPTIONAL when b_submission=true (value is supplied at evaluation time by user submission).",
+                          "description": "The actual value data. Format depends on `value_type`:\nTYPE-CLASS SPLIT (F03): scalar value_types (Bool/Address/String/U8..U256) take ONE scalar value (boolean/address string/name/string/number); Vec* value_types (VecBool/VecAddress/VecString/VecU8..VecU256) take a REAL JSON ARRAY of elements — a string (even a Python-repr-looking \"[a, b]\") fails serialization with 'value_type=10 (array) requires a array value'. The historical 'static values must be strings' rule applied to SCALARS only.\n• Bool: true/false (boolean)\n• Address (CRITICAL — string 'Address' is NOT a placeholder): a hex address (e.g. '0x1234...'), a LocalMark name (e.g. 'my-service' — resolved to address at evaluation time), an AccountOrMark_Address object (e.g. {name_or_address:'my-service'}), or system shorthand ('0xaaa' = EntityLinker, '0xaab' = EntityRegistrar). The literal string 'Address' itself is INVALID — it would be treated as a non-existent LocalMark name and fail. Example: value='0x2::wow::WOW<address>' or value='my-permission'.\n• String: any string\n• U8/U16/U32/U64/U128/U256: number or numeric string (e.g. 42 or '42')\n• Vec* types: arrays of the corresponding element type\nPROTOCOL SLOT 254 (once, b_submission Address): the submitted value MUST be the ORDER address bound to the operation (payment_info.for_object) — NEVER the Progress/Machine/Arb object the call is advancing. The chain asserts submission(254) == the order address; a Progress address fails the task binding and the forward aborts.\nREQUIRED when b_submission=false. OPTIONAL when b_submission=true (value is supplied at evaluation time by user submission).",
                           "anyOf": [
                             {
                               "type": "boolean"
@@ -5166,7 +5229,7 @@
                       "additionalProperties": false,
                       "description": "Guard table item (QUERY/OUTPUT form — includes auto-derived object_type field)"
                     },
-                    "description": "User-submitted data matching the Guard's required fields. Relation: structure must match the Guard table's column definitions. Example: [{field:'delivery_proof', value:'Qm...'}]"
+                    "description": "User-submitted data matching the Guard's required fields. Relation: structure must match the Guard table's column definitions. Example: [{field:'delivery_proof', value:'Qm...'}] PROTOCOL SLOT 254: when the Guard table has identifier 254, submit the ORDER address bound to the operation (payment_info.for_object) — never the Progress/Machine object being advanced."
                   }
                 },
                 "required": [
@@ -5176,7 +5239,7 @@
                 "additionalProperties": false,
                 "description": "One Guard's submission data: the Guard to verify plus the user-provided data that satisfies its requirements."
               },
-              "description": "User-submitted data for each Guard. Relation: one entry per Guard in the guard array; fill submission fields and resubmit via call_with_submission. PLACEMENT: this `submission` field is at the SAME level as `data` and `env` in the operation input — NOT inside `data.data`. Example structure: {tool:'onchain_operations', data:{operation_type:'order', data:{object:'my_order', progress:{...}}}, submission:{type:'submission', guard:[...], submission:[...]}}"
+              "description": "User-submitted data for each Guard. Relation: one entry per Guard in the guard array; fill submission fields and resubmit via call_with_submission. SLOT COMPLETENESS (T8 az3-mn-261013): the submission array must cover EVERY b_submission slot in the Guard's table, not just one — a Guard whose table has multiple b_submission entries (e.g. slot 0 + slot 254) aborts on-chain with 'Submission ID not found' (abort 11) when any b_submission slot has no matching entry. Check the Guard's table (each entry with b_submission needs its identifier filled) and include ALL of them. PLACEMENT: this `submission` field is at the SAME level as `data` and `env` in the operation input — NOT inside `data.data`. Example structure: {tool:'onchain_operations', data:{operation_type:'order', data:{object:'my_order', progress:{...}}}, submission:{type:'submission', guard:[...], submission:[...]}}"
             }
           },
           "required": [
@@ -5284,7 +5347,7 @@
               "description": "Description string (max 4000 bcs characters)"
             },
             "policies": {
-              "description": "Write and read permission rules governing the data records of this Repository.",
+              "description": "Write and read permission rules governing the data records of this Repository. NOTE the PLURAL field name here — the Permission object's equivalent field is singular 'policy' (set/add/remove ops); do not carry one spelling to the other.",
               "oneOf": [
                 {
                   "type": "object",
@@ -5401,6 +5464,7 @@
                             "description": "Source of data ID when writing data. If None, user must specify the ID separately; if Clock, data ID is current timestamp; if Signer, data ID is user ID."
                           },
                           "value_type": {
+                            "description": "Type of data value when writing data. Note: 19 (Value) is an internal dynamic type for guard expressions and is NOT accepted here.",
                             "anyOf": [
                               {
                                 "type": "number",
@@ -5687,8 +5751,7 @@
                                 "const": "vecvecu8",
                                 "description": "vecvecu8"
                               }
-                            ],
-                            "description": "Type of data value when writing data. Note: 19 (Value) is an internal dynamic type for guard expressions and is NOT accepted here."
+                            ]
                           }
                         },
                         "required": [
@@ -6011,6 +6074,7 @@
               "description": "Unwrap CoinWrapper objects and other objects received by this Repository object and send them to the owner of its Permission object.\n\nACCEPTED FORMATS (F-06 unified receive operation block):\n• 'recently' (string literal) — auto-query and receive ALL recently received objects.\n  Use this for the common case: \"withdraw everything the object has received.\"\n  Example: receive: 'recently'\n• ReceivedNormal[] (array) — explicit list of received objects to unwrap.\n  Use this when you want to receive specific objects only (not all).\n  Example: receive: [{id: '0xobj1', type: '0x2::coin::Coin<0x2::wow::WOW>'}]\n• ReceivedBalance ({token_type, balance, received: [{id, balance, payment}]}) —\n  receive a balance record from a specific Payment/payer.\n  Use this for precise balance targeting (advanced — usually after querying\n  the object's received history via query_received).\n  Example: receive: {token_type: '0x2::wow::WOW', balance: 1000000, received: [{id: '0xrec1', balance: 1000000, payment: '0xpay1'}]}\nANTI-PATTERN: do NOT wrap in {result: ...} — pass the value directly.",
               "anyOf": [
                 {
+                  "minItems": 1,
                   "type": "array",
                   "items": {
                     "type": "object",
@@ -6289,6 +6353,7 @@
                           "description": "Whether this table item's value is submitted dynamically at Guard trigger time (alloc_by_guard call). \n\ntrue = value is submitted by the caller when triggering the Guard. Use for runtime-context-dependent values like order address, user address. The 'value' field is ignored when b_submission=true; the caller must provide it via submissions[]. \n\nfalse = value is static, set at Guard creation time. Use for values known when the Guard is created: expected node names, expected merchant address, expected service address. The 'value' field must be populated and will be stored on-chain permanently. \n\nRule of thumb: if the value is the SAME for all future Guard triggers, use false. If the value DIFFERS per trigger (e.g., which order to release funds for), use true."
                         },
                         "value_type": {
+                          "description": "Type of the value stored in `value`. One of: Bool(0), Address(1), String(2), U8(3), U16(4), U32(5), U64(6), U128(7), U256(8), VecBool(9), VecAddress(10), VecString(11), VecU8(12), VecU16(13), VecU32(14), VecU64(15), VecU128(16), VecU256(17), VecVecU8(18). When value_type=Address (1), the `value` field accepts a hex address string, a LocalMark name, or an AccountOrMark_Address object — see `value` field description for details.",
                           "anyOf": [
                             {
                               "type": "number",
@@ -6575,11 +6640,10 @@
                               "const": "vecvecu8",
                               "description": "vecvecu8"
                             }
-                          ],
-                          "description": "Type of the value stored in `value`. One of: Bool(0), Address(1), String(2), U8(3), U16(4), U32(5), U64(6), U128(7), U256(8), VecBool(9), VecAddress(10), VecString(11), VecU8(12), VecU16(13), VecU32(14), VecU64(15), VecU128(16), VecU256(17), VecVecU8(18). When value_type=Address (1), the `value` field accepts a hex address string, a LocalMark name, or an AccountOrMark_Address object — see `value` field description for details."
+                          ]
                         },
                         "value": {
-                          "description": "The actual value data. Format depends on `value_type`:\n• Bool: true/false (boolean)\n• Address (CRITICAL — string 'Address' is NOT a placeholder): a hex address (e.g. '0x1234...'), a LocalMark name (e.g. 'my-service' — resolved to address at evaluation time), an AccountOrMark_Address object (e.g. {name_or_address:'my-service'}), or system shorthand ('0xaaa' = EntityLinker, '0xaab' = EntityRegistrar). The literal string 'Address' itself is INVALID — it would be treated as a non-existent LocalMark name and fail. Example: value='0x2::wow::WOW<address>' or value='my-permission'.\n• String: any string\n• U8/U16/U32/U64/U128/U256: number or numeric string (e.g. 42 or '42')\n• Vec* types: arrays of the corresponding element type\nREQUIRED when b_submission=false. OPTIONAL when b_submission=true (value is supplied at evaluation time by user submission).",
+                          "description": "The actual value data. Format depends on `value_type`:\nTYPE-CLASS SPLIT (F03): scalar value_types (Bool/Address/String/U8..U256) take ONE scalar value (boolean/address string/name/string/number); Vec* value_types (VecBool/VecAddress/VecString/VecU8..VecU256) take a REAL JSON ARRAY of elements — a string (even a Python-repr-looking \"[a, b]\") fails serialization with 'value_type=10 (array) requires a array value'. The historical 'static values must be strings' rule applied to SCALARS only.\n• Bool: true/false (boolean)\n• Address (CRITICAL — string 'Address' is NOT a placeholder): a hex address (e.g. '0x1234...'), a LocalMark name (e.g. 'my-service' — resolved to address at evaluation time), an AccountOrMark_Address object (e.g. {name_or_address:'my-service'}), or system shorthand ('0xaaa' = EntityLinker, '0xaab' = EntityRegistrar). The literal string 'Address' itself is INVALID — it would be treated as a non-existent LocalMark name and fail. Example: value='0x2::wow::WOW<address>' or value='my-permission'.\n• String: any string\n• U8/U16/U32/U64/U128/U256: number or numeric string (e.g. 42 or '42')\n• Vec* types: arrays of the corresponding element type\nPROTOCOL SLOT 254 (once, b_submission Address): the submitted value MUST be the ORDER address bound to the operation (payment_info.for_object) — NEVER the Progress/Machine/Arb object the call is advancing. The chain asserts submission(254) == the order address; a Progress address fails the task binding and the forward aborts.\nREQUIRED when b_submission=false. OPTIONAL when b_submission=true (value is supplied at evaluation time by user submission).",
                           "anyOf": [
                             {
                               "type": "boolean"
@@ -6738,7 +6802,7 @@
                       "additionalProperties": false,
                       "description": "Guard table item (QUERY/OUTPUT form — includes auto-derived object_type field)"
                     },
-                    "description": "User-submitted data matching the Guard's required fields. Relation: structure must match the Guard table's column definitions. Example: [{field:'delivery_proof', value:'Qm...'}]"
+                    "description": "User-submitted data matching the Guard's required fields. Relation: structure must match the Guard table's column definitions. Example: [{field:'delivery_proof', value:'Qm...'}] PROTOCOL SLOT 254: when the Guard table has identifier 254, submit the ORDER address bound to the operation (payment_info.for_object) — never the Progress/Machine object being advanced."
                   }
                 },
                 "required": [
@@ -6748,7 +6812,7 @@
                 "additionalProperties": false,
                 "description": "One Guard's submission data: the Guard to verify plus the user-provided data that satisfies its requirements."
               },
-              "description": "User-submitted data for each Guard. Relation: one entry per Guard in the guard array; fill submission fields and resubmit via call_with_submission. PLACEMENT: this `submission` field is at the SAME level as `data` and `env` in the operation input — NOT inside `data.data`. Example structure: {tool:'onchain_operations', data:{operation_type:'order', data:{object:'my_order', progress:{...}}}, submission:{type:'submission', guard:[...], submission:[...]}}"
+              "description": "User-submitted data for each Guard. Relation: one entry per Guard in the guard array; fill submission fields and resubmit via call_with_submission. SLOT COMPLETENESS (T8 az3-mn-261013): the submission array must cover EVERY b_submission slot in the Guard's table, not just one — a Guard whose table has multiple b_submission entries (e.g. slot 0 + slot 254) aborts on-chain with 'Submission ID not found' (abort 11) when any b_submission slot has no matching entry. Check the Guard's table (each entry with b_submission needs its identifier filled) and include ALL of them. PLACEMENT: this `submission` field is at the SAME level as `data` and `env` in the operation input — NOT inside `data.data`. Example structure: {tool:'onchain_operations', data:{operation_type:'order', data:{object:'my_order', progress:{...}}}, submission:{type:'submission', guard:[...], submission:[...]}}"
             }
           },
           "required": [
@@ -6868,8 +6932,8 @@
                   "description": "Order ID or name."
                 },
                 "description": {
-                  "description": "Dispute description.",
-                  "type": "string"
+                  "type": "string",
+                  "description": "Dispute description. REQUIRED and NON-EMPTY on-chain (the dispute call aborts when it is empty; max 4000 bytes). State what the dispute is about — it is permanent on-chain case context."
                 },
                 "proposition": {
                   "type": "array",
@@ -6946,6 +7010,7 @@
               },
               "required": [
                 "order",
+                "description",
                 "proposition",
                 "fee"
               ],
@@ -7047,7 +7112,7 @@
                   "description": "Proposition INDICES (0-based, u8) the voter agrees with — e.g. [0] votes for the first proposition. Re-voting REPLACES the voter's previous vote (old weight removed, new applied). Out-of-range index aborts with E_PROPOSITION_NOT_FOUND."
                 },
                 "voting_guard": {
-                  "description": "Voting Guard for WEIGHTED voting. DECISION RULE (contract-enforced):\n(A) Arbitration HAS a non-empty voting_guard list → plain vote is REJECTED on-chain with E_VOTING_GUARD_USED (abort 2). You MUST provide this field set to ONE guard from the Arbitration's voting_guard list, and fill its submission: when the guard's vote_weight is {GuardIdentifier: pos}, the number you submit at that table position IS the vote weight; when it is {FixedValue: v}, no submission is needed (weight is fixed). The SDK then routes the call through the Guard-weighted vote path via a Passport. Omitting this field on such an Arbitration now fails fast pre-call with the list of valid guards.\n(B) Arbitration has NO voting_guard → omit this field; the SDK uses plain vote (weight 1, permission-gated), or the passport-gated vote path when other guards require it.\nBOTH (B) paths still assert ARBITRATION_VOTE=358 on-chain (the SDK now pre-checks 358 and fails fast with 'Missing permissions 358'). An empty voting_guard means EQUAL WEIGHT (1 person 1 vote), NOT permissionless — there is no permission-free vote entry. For genuinely open voting, configure a public voting_guard (op=\"set\") and vote via path (A): the weighted passport path authorizes by the guard and does NOT require 358.\nThe Guard (when provided) MUST be in the Arbitration's voting_guard list — an unlisted guard aborts with E_VOTING_GUARD_NOT_FOUND.",
+                  "description": "Voting Guard for WEIGHTED voting. DECISION RULE (contract-enforced):\n(A) Arbitration HAS a non-empty voting_guard list → plain vote is REJECTED on-chain with E_VOTING_GUARD_USED (abort 2). You MUST provide this field set to ONE guard from the Arbitration's voting_guard list, and fill its submission: when the guard's vote_weight is {GuardIdentifier: pos}, the number you submit at that table position IS the vote weight; when it is {FixedValue: v}, no submission is needed (weight is fixed). The SDK then routes the call through the Guard-weighted vote path via a Passport. Omitting this field on such an Arbitration now fails fast pre-call with the list of valid guards.\n(B) Arbitration has NO voting_guard → omit this field; the SDK uses plain vote (weight 1, permission-gated), or the passport-gated vote path when other guards require it.\nBOTH (B) paths still assert ARBITRATION_VOTE=358 on-chain (the SDK now pre-checks 358 and fails fast with 'Missing permissions 358'). An empty voting_guard means EQUAL WEIGHT (1 person 1 vote), NOT permissionless — there is no permission-free vote entry. For genuinely open voting, configure a public voting_guard (op=\"set\") and vote via path (A): the weighted passport path authorizes by the guard and does NOT require 358.\nThe Guard (when provided) MUST be in the Arbitration's voting_guard list — an unlisted guard aborts with E_VOTING_GUARD_NOT_FOUND. WEIGHTED REVOTE RESIDUE (az3-mn-261013, arb.move:315-320 source comment): under weighted voting a re-vote is NOT clean — re-voting with a LARGER weight than the previous one aborts on underflow (re-vote DoS surface); re-voting with a SMALLER weight succeeds but leaves phantom votes behind. Treat the first weighted vote as final when possible.",
                   "type": "string"
                 }
               },
@@ -7091,7 +7156,8 @@
                 "indemnity": {
                   "type": "integer",
                   "minimum": 0,
-                  "maximum": 9007199254740991
+                  "maximum": 9007199254740991,
+                  "description": "Awarded compensation in integer smallest units (0 = a denial ruling that pays nothing). Hard-capped at 3x the disputed order amount on chain. Funded from the Service's merchant-prepaid compensation fund — an award larger than the fund balance aborts the later payout until the merchant tops up. A non-zero award consumes the order's one-shot compensation latch: refund rails gated on 'no award paid' can never fire afterwards, so pair them with a merchant-recovery rail at build time."
                 }
               },
               "required": [
@@ -7340,6 +7406,7 @@
               "description": "Unwrap CoinWrapper objects and other objects received by this Arbitration object and send them to the owner of its Permission object.\n\nACCEPTED FORMATS (F-06 unified receive operation block):\n• 'recently' (string literal) — auto-query and receive ALL recently received objects.\n  Use this for the common case: \"withdraw everything the object has received.\"\n  Example: receive: 'recently'\n• ReceivedNormal[] (array) — explicit list of received objects to unwrap.\n  Use this when you want to receive specific objects only (not all).\n  Example: receive: [{id: '0xobj1', type: '0x2::coin::Coin<0x2::wow::WOW>'}]\n• ReceivedBalance ({token_type, balance, received: [{id, balance, payment}]}) —\n  receive a balance record from a specific Payment/payer.\n  Use this for precise balance targeting (advanced — usually after querying\n  the object's received history via query_received).\n  Example: receive: {token_type: '0x2::wow::WOW', balance: 1000000, received: [{id: '0xrec1', balance: 1000000, payment: '0xpay1'}]}\nANTI-PATTERN: do NOT wrap in {result: ...} — pass the value directly.",
               "anyOf": [
                 {
+                  "minItems": 1,
                   "type": "array",
                   "items": {
                     "type": "object",
@@ -7618,6 +7685,7 @@
                           "description": "Whether this table item's value is submitted dynamically at Guard trigger time (alloc_by_guard call). \n\ntrue = value is submitted by the caller when triggering the Guard. Use for runtime-context-dependent values like order address, user address. The 'value' field is ignored when b_submission=true; the caller must provide it via submissions[]. \n\nfalse = value is static, set at Guard creation time. Use for values known when the Guard is created: expected node names, expected merchant address, expected service address. The 'value' field must be populated and will be stored on-chain permanently. \n\nRule of thumb: if the value is the SAME for all future Guard triggers, use false. If the value DIFFERS per trigger (e.g., which order to release funds for), use true."
                         },
                         "value_type": {
+                          "description": "Type of the value stored in `value`. One of: Bool(0), Address(1), String(2), U8(3), U16(4), U32(5), U64(6), U128(7), U256(8), VecBool(9), VecAddress(10), VecString(11), VecU8(12), VecU16(13), VecU32(14), VecU64(15), VecU128(16), VecU256(17), VecVecU8(18). When value_type=Address (1), the `value` field accepts a hex address string, a LocalMark name, or an AccountOrMark_Address object — see `value` field description for details.",
                           "anyOf": [
                             {
                               "type": "number",
@@ -7904,11 +7972,10 @@
                               "const": "vecvecu8",
                               "description": "vecvecu8"
                             }
-                          ],
-                          "description": "Type of the value stored in `value`. One of: Bool(0), Address(1), String(2), U8(3), U16(4), U32(5), U64(6), U128(7), U256(8), VecBool(9), VecAddress(10), VecString(11), VecU8(12), VecU16(13), VecU32(14), VecU64(15), VecU128(16), VecU256(17), VecVecU8(18). When value_type=Address (1), the `value` field accepts a hex address string, a LocalMark name, or an AccountOrMark_Address object — see `value` field description for details."
+                          ]
                         },
                         "value": {
-                          "description": "The actual value data. Format depends on `value_type`:\n• Bool: true/false (boolean)\n• Address (CRITICAL — string 'Address' is NOT a placeholder): a hex address (e.g. '0x1234...'), a LocalMark name (e.g. 'my-service' — resolved to address at evaluation time), an AccountOrMark_Address object (e.g. {name_or_address:'my-service'}), or system shorthand ('0xaaa' = EntityLinker, '0xaab' = EntityRegistrar). The literal string 'Address' itself is INVALID — it would be treated as a non-existent LocalMark name and fail. Example: value='0x2::wow::WOW<address>' or value='my-permission'.\n• String: any string\n• U8/U16/U32/U64/U128/U256: number or numeric string (e.g. 42 or '42')\n• Vec* types: arrays of the corresponding element type\nREQUIRED when b_submission=false. OPTIONAL when b_submission=true (value is supplied at evaluation time by user submission).",
+                          "description": "The actual value data. Format depends on `value_type`:\nTYPE-CLASS SPLIT (F03): scalar value_types (Bool/Address/String/U8..U256) take ONE scalar value (boolean/address string/name/string/number); Vec* value_types (VecBool/VecAddress/VecString/VecU8..VecU256) take a REAL JSON ARRAY of elements — a string (even a Python-repr-looking \"[a, b]\") fails serialization with 'value_type=10 (array) requires a array value'. The historical 'static values must be strings' rule applied to SCALARS only.\n• Bool: true/false (boolean)\n• Address (CRITICAL — string 'Address' is NOT a placeholder): a hex address (e.g. '0x1234...'), a LocalMark name (e.g. 'my-service' — resolved to address at evaluation time), an AccountOrMark_Address object (e.g. {name_or_address:'my-service'}), or system shorthand ('0xaaa' = EntityLinker, '0xaab' = EntityRegistrar). The literal string 'Address' itself is INVALID — it would be treated as a non-existent LocalMark name and fail. Example: value='0x2::wow::WOW<address>' or value='my-permission'.\n• String: any string\n• U8/U16/U32/U64/U128/U256: number or numeric string (e.g. 42 or '42')\n• Vec* types: arrays of the corresponding element type\nPROTOCOL SLOT 254 (once, b_submission Address): the submitted value MUST be the ORDER address bound to the operation (payment_info.for_object) — NEVER the Progress/Machine/Arb object the call is advancing. The chain asserts submission(254) == the order address; a Progress address fails the task binding and the forward aborts.\nREQUIRED when b_submission=false. OPTIONAL when b_submission=true (value is supplied at evaluation time by user submission).",
                           "anyOf": [
                             {
                               "type": "boolean"
@@ -8067,7 +8134,7 @@
                       "additionalProperties": false,
                       "description": "Guard table item (QUERY/OUTPUT form — includes auto-derived object_type field)"
                     },
-                    "description": "User-submitted data matching the Guard's required fields. Relation: structure must match the Guard table's column definitions. Example: [{field:'delivery_proof', value:'Qm...'}]"
+                    "description": "User-submitted data matching the Guard's required fields. Relation: structure must match the Guard table's column definitions. Example: [{field:'delivery_proof', value:'Qm...'}] PROTOCOL SLOT 254: when the Guard table has identifier 254, submit the ORDER address bound to the operation (payment_info.for_object) — never the Progress/Machine object being advanced."
                   }
                 },
                 "required": [
@@ -8077,7 +8144,7 @@
                 "additionalProperties": false,
                 "description": "One Guard's submission data: the Guard to verify plus the user-provided data that satisfies its requirements."
               },
-              "description": "User-submitted data for each Guard. Relation: one entry per Guard in the guard array; fill submission fields and resubmit via call_with_submission. PLACEMENT: this `submission` field is at the SAME level as `data` and `env` in the operation input — NOT inside `data.data`. Example structure: {tool:'onchain_operations', data:{operation_type:'order', data:{object:'my_order', progress:{...}}}, submission:{type:'submission', guard:[...], submission:[...]}}"
+              "description": "User-submitted data for each Guard. Relation: one entry per Guard in the guard array; fill submission fields and resubmit via call_with_submission. SLOT COMPLETENESS (T8 az3-mn-261013): the submission array must cover EVERY b_submission slot in the Guard's table, not just one — a Guard whose table has multiple b_submission entries (e.g. slot 0 + slot 254) aborts on-chain with 'Submission ID not found' (abort 11) when any b_submission slot has no matching entry. Check the Guard's table (each entry with b_submission needs its identifier filled) and include ALL of them. PLACEMENT: this `submission` field is at the SAME level as `data` and `env` in the operation input — NOT inside `data.data`. Example structure: {tool:'onchain_operations', data:{operation_type:'order', data:{object:'my_order', progress:{...}}}, submission:{type:'submission', guard:[...], submission:[...]}}"
             }
           },
           "required": [
@@ -8221,7 +8288,7 @@
                           "at"
                         ],
                         "additionalProperties": false,
-                        "description": "An instant messaging contact entry with address and optional description"
+                        "description": "An instant messaging contact entry (at + optional description). NO status field by design: on chain, Im.status is self-declared — each member sets its OWN status (your own via data.my_status; other members via their own contact-status update), and creating an entry always starts it with an empty status."
                       },
                       "description": "List of IM contacts to add"
                     }
@@ -8259,7 +8326,7 @@
                           "at"
                         ],
                         "additionalProperties": false,
-                        "description": "An instant messaging contact entry with address and optional description"
+                        "description": "An instant messaging contact entry (at + optional description). NO status field by design: on chain, Im.status is self-declared — each member sets its OWN status (your own via data.my_status; other members via their own contact-status update), and creating an entry always starts it with an empty status."
                       },
                       "description": "Complete list of IM contacts to set"
                     }
@@ -8316,6 +8383,7 @@
               "description": "Receive objects sent to this Contact object and unwrap them to the permission owner.\n\nACCEPTED FORMATS (F-06 unified receive operation block):\n• 'recently' (string literal) — auto-query and receive ALL recently received objects.\n  Use this for the common case: \"withdraw everything the object has received.\"\n  Example: receive: 'recently'\n• ReceivedNormal[] (array) — explicit list of received objects to unwrap.\n  Use this when you want to receive specific objects only (not all).\n  Example: receive: [{id: '0xobj1', type: '0x2::coin::Coin<0x2::wow::WOW>'}]\n• ReceivedBalance ({token_type, balance, received: [{id, balance, payment}]}) —\n  receive a balance record from a specific Payment/payer.\n  Use this for precise balance targeting (advanced — usually after querying\n  the object's received history via query_received).\n  Example: receive: {token_type: '0x2::wow::WOW', balance: 1000000, received: [{id: '0xrec1', balance: 1000000, payment: '0xpay1'}]}\nANTI-PATTERN: do NOT wrap in {result: ...} — pass the value directly.",
               "anyOf": [
                 {
+                  "minItems": 1,
                   "type": "array",
                   "items": {
                     "type": "object",
@@ -8582,6 +8650,7 @@
                           "description": "Whether this table item's value is submitted dynamically at Guard trigger time (alloc_by_guard call). \n\ntrue = value is submitted by the caller when triggering the Guard. Use for runtime-context-dependent values like order address, user address. The 'value' field is ignored when b_submission=true; the caller must provide it via submissions[]. \n\nfalse = value is static, set at Guard creation time. Use for values known when the Guard is created: expected node names, expected merchant address, expected service address. The 'value' field must be populated and will be stored on-chain permanently. \n\nRule of thumb: if the value is the SAME for all future Guard triggers, use false. If the value DIFFERS per trigger (e.g., which order to release funds for), use true."
                         },
                         "value_type": {
+                          "description": "Type of the value stored in `value`. One of: Bool(0), Address(1), String(2), U8(3), U16(4), U32(5), U64(6), U128(7), U256(8), VecBool(9), VecAddress(10), VecString(11), VecU8(12), VecU16(13), VecU32(14), VecU64(15), VecU128(16), VecU256(17), VecVecU8(18). When value_type=Address (1), the `value` field accepts a hex address string, a LocalMark name, or an AccountOrMark_Address object — see `value` field description for details.",
                           "anyOf": [
                             {
                               "type": "number",
@@ -8868,11 +8937,10 @@
                               "const": "vecvecu8",
                               "description": "vecvecu8"
                             }
-                          ],
-                          "description": "Type of the value stored in `value`. One of: Bool(0), Address(1), String(2), U8(3), U16(4), U32(5), U64(6), U128(7), U256(8), VecBool(9), VecAddress(10), VecString(11), VecU8(12), VecU16(13), VecU32(14), VecU64(15), VecU128(16), VecU256(17), VecVecU8(18). When value_type=Address (1), the `value` field accepts a hex address string, a LocalMark name, or an AccountOrMark_Address object — see `value` field description for details."
+                          ]
                         },
                         "value": {
-                          "description": "The actual value data. Format depends on `value_type`:\n• Bool: true/false (boolean)\n• Address (CRITICAL — string 'Address' is NOT a placeholder): a hex address (e.g. '0x1234...'), a LocalMark name (e.g. 'my-service' — resolved to address at evaluation time), an AccountOrMark_Address object (e.g. {name_or_address:'my-service'}), or system shorthand ('0xaaa' = EntityLinker, '0xaab' = EntityRegistrar). The literal string 'Address' itself is INVALID — it would be treated as a non-existent LocalMark name and fail. Example: value='0x2::wow::WOW<address>' or value='my-permission'.\n• String: any string\n• U8/U16/U32/U64/U128/U256: number or numeric string (e.g. 42 or '42')\n• Vec* types: arrays of the corresponding element type\nREQUIRED when b_submission=false. OPTIONAL when b_submission=true (value is supplied at evaluation time by user submission).",
+                          "description": "The actual value data. Format depends on `value_type`:\nTYPE-CLASS SPLIT (F03): scalar value_types (Bool/Address/String/U8..U256) take ONE scalar value (boolean/address string/name/string/number); Vec* value_types (VecBool/VecAddress/VecString/VecU8..VecU256) take a REAL JSON ARRAY of elements — a string (even a Python-repr-looking \"[a, b]\") fails serialization with 'value_type=10 (array) requires a array value'. The historical 'static values must be strings' rule applied to SCALARS only.\n• Bool: true/false (boolean)\n• Address (CRITICAL — string 'Address' is NOT a placeholder): a hex address (e.g. '0x1234...'), a LocalMark name (e.g. 'my-service' — resolved to address at evaluation time), an AccountOrMark_Address object (e.g. {name_or_address:'my-service'}), or system shorthand ('0xaaa' = EntityLinker, '0xaab' = EntityRegistrar). The literal string 'Address' itself is INVALID — it would be treated as a non-existent LocalMark name and fail. Example: value='0x2::wow::WOW<address>' or value='my-permission'.\n• String: any string\n• U8/U16/U32/U64/U128/U256: number or numeric string (e.g. 42 or '42')\n• Vec* types: arrays of the corresponding element type\nPROTOCOL SLOT 254 (once, b_submission Address): the submitted value MUST be the ORDER address bound to the operation (payment_info.for_object) — NEVER the Progress/Machine/Arb object the call is advancing. The chain asserts submission(254) == the order address; a Progress address fails the task binding and the forward aborts.\nREQUIRED when b_submission=false. OPTIONAL when b_submission=true (value is supplied at evaluation time by user submission).",
                           "anyOf": [
                             {
                               "type": "boolean"
@@ -9031,7 +9099,7 @@
                       "additionalProperties": false,
                       "description": "Guard table item (QUERY/OUTPUT form — includes auto-derived object_type field)"
                     },
-                    "description": "User-submitted data matching the Guard's required fields. Relation: structure must match the Guard table's column definitions. Example: [{field:'delivery_proof', value:'Qm...'}]"
+                    "description": "User-submitted data matching the Guard's required fields. Relation: structure must match the Guard table's column definitions. Example: [{field:'delivery_proof', value:'Qm...'}] PROTOCOL SLOT 254: when the Guard table has identifier 254, submit the ORDER address bound to the operation (payment_info.for_object) — never the Progress/Machine object being advanced."
                   }
                 },
                 "required": [
@@ -9041,7 +9109,7 @@
                 "additionalProperties": false,
                 "description": "One Guard's submission data: the Guard to verify plus the user-provided data that satisfies its requirements."
               },
-              "description": "User-submitted data for each Guard. Relation: one entry per Guard in the guard array; fill submission fields and resubmit via call_with_submission. PLACEMENT: this `submission` field is at the SAME level as `data` and `env` in the operation input — NOT inside `data.data`. Example structure: {tool:'onchain_operations', data:{operation_type:'order', data:{object:'my_order', progress:{...}}}, submission:{type:'submission', guard:[...], submission:[...]}}"
+              "description": "User-submitted data for each Guard. Relation: one entry per Guard in the guard array; fill submission fields and resubmit via call_with_submission. SLOT COMPLETENESS (T8 az3-mn-261013): the submission array must cover EVERY b_submission slot in the Guard's table, not just one — a Guard whose table has multiple b_submission entries (e.g. slot 0 + slot 254) aborts on-chain with 'Submission ID not found' (abort 11) when any b_submission slot has no matching entry. Check the Guard's table (each entry with b_submission needs its identifier filled) and include ALL of them. PLACEMENT: this `submission` field is at the SAME level as `data` and `env` in the operation input — NOT inside `data.data`. Example structure: {tool:'onchain_operations', data:{operation_type:'order', data:{object:'my_order', progress:{...}}}, submission:{type:'submission', guard:[...], submission:[...]}}"
             }
           },
           "required": [
@@ -9809,6 +9877,7 @@
               "description": "Unwrap CoinWrapper objects and other objects received by this Treasury object and send them to the owner of its Permission object.\n\nACCEPTED FORMATS (F-06 unified receive operation block):\n• 'recently' (string literal) — auto-query and receive ALL recently received objects.\n  Use this for the common case: \"withdraw everything the object has received.\"\n  Example: receive: 'recently'\n• ReceivedNormal[] (array) — explicit list of received objects to unwrap.\n  Use this when you want to receive specific objects only (not all).\n  Example: receive: [{id: '0xobj1', type: '0x2::coin::Coin<0x2::wow::WOW>'}]\n• ReceivedBalance ({token_type, balance, received: [{id, balance, payment}]}) —\n  receive a balance record from a specific Payment/payer.\n  Use this for precise balance targeting (advanced — usually after querying\n  the object's received history via query_received).\n  Example: receive: {token_type: '0x2::wow::WOW', balance: 1000000, received: [{id: '0xrec1', balance: 1000000, payment: '0xpay1'}]}\nANTI-PATTERN: do NOT wrap in {result: ...} — pass the value directly.",
               "anyOf": [
                 {
+                  "minItems": 1,
                   "type": "array",
                   "items": {
                     "type": "object",
@@ -10087,6 +10156,7 @@
                           "description": "Whether this table item's value is submitted dynamically at Guard trigger time (alloc_by_guard call). \n\ntrue = value is submitted by the caller when triggering the Guard. Use for runtime-context-dependent values like order address, user address. The 'value' field is ignored when b_submission=true; the caller must provide it via submissions[]. \n\nfalse = value is static, set at Guard creation time. Use for values known when the Guard is created: expected node names, expected merchant address, expected service address. The 'value' field must be populated and will be stored on-chain permanently. \n\nRule of thumb: if the value is the SAME for all future Guard triggers, use false. If the value DIFFERS per trigger (e.g., which order to release funds for), use true."
                         },
                         "value_type": {
+                          "description": "Type of the value stored in `value`. One of: Bool(0), Address(1), String(2), U8(3), U16(4), U32(5), U64(6), U128(7), U256(8), VecBool(9), VecAddress(10), VecString(11), VecU8(12), VecU16(13), VecU32(14), VecU64(15), VecU128(16), VecU256(17), VecVecU8(18). When value_type=Address (1), the `value` field accepts a hex address string, a LocalMark name, or an AccountOrMark_Address object — see `value` field description for details.",
                           "anyOf": [
                             {
                               "type": "number",
@@ -10373,11 +10443,10 @@
                               "const": "vecvecu8",
                               "description": "vecvecu8"
                             }
-                          ],
-                          "description": "Type of the value stored in `value`. One of: Bool(0), Address(1), String(2), U8(3), U16(4), U32(5), U64(6), U128(7), U256(8), VecBool(9), VecAddress(10), VecString(11), VecU8(12), VecU16(13), VecU32(14), VecU64(15), VecU128(16), VecU256(17), VecVecU8(18). When value_type=Address (1), the `value` field accepts a hex address string, a LocalMark name, or an AccountOrMark_Address object — see `value` field description for details."
+                          ]
                         },
                         "value": {
-                          "description": "The actual value data. Format depends on `value_type`:\n• Bool: true/false (boolean)\n• Address (CRITICAL — string 'Address' is NOT a placeholder): a hex address (e.g. '0x1234...'), a LocalMark name (e.g. 'my-service' — resolved to address at evaluation time), an AccountOrMark_Address object (e.g. {name_or_address:'my-service'}), or system shorthand ('0xaaa' = EntityLinker, '0xaab' = EntityRegistrar). The literal string 'Address' itself is INVALID — it would be treated as a non-existent LocalMark name and fail. Example: value='0x2::wow::WOW<address>' or value='my-permission'.\n• String: any string\n• U8/U16/U32/U64/U128/U256: number or numeric string (e.g. 42 or '42')\n• Vec* types: arrays of the corresponding element type\nREQUIRED when b_submission=false. OPTIONAL when b_submission=true (value is supplied at evaluation time by user submission).",
+                          "description": "The actual value data. Format depends on `value_type`:\nTYPE-CLASS SPLIT (F03): scalar value_types (Bool/Address/String/U8..U256) take ONE scalar value (boolean/address string/name/string/number); Vec* value_types (VecBool/VecAddress/VecString/VecU8..VecU256) take a REAL JSON ARRAY of elements — a string (even a Python-repr-looking \"[a, b]\") fails serialization with 'value_type=10 (array) requires a array value'. The historical 'static values must be strings' rule applied to SCALARS only.\n• Bool: true/false (boolean)\n• Address (CRITICAL — string 'Address' is NOT a placeholder): a hex address (e.g. '0x1234...'), a LocalMark name (e.g. 'my-service' — resolved to address at evaluation time), an AccountOrMark_Address object (e.g. {name_or_address:'my-service'}), or system shorthand ('0xaaa' = EntityLinker, '0xaab' = EntityRegistrar). The literal string 'Address' itself is INVALID — it would be treated as a non-existent LocalMark name and fail. Example: value='0x2::wow::WOW<address>' or value='my-permission'.\n• String: any string\n• U8/U16/U32/U64/U128/U256: number or numeric string (e.g. 42 or '42')\n• Vec* types: arrays of the corresponding element type\nPROTOCOL SLOT 254 (once, b_submission Address): the submitted value MUST be the ORDER address bound to the operation (payment_info.for_object) — NEVER the Progress/Machine/Arb object the call is advancing. The chain asserts submission(254) == the order address; a Progress address fails the task binding and the forward aborts.\nREQUIRED when b_submission=false. OPTIONAL when b_submission=true (value is supplied at evaluation time by user submission).",
                           "anyOf": [
                             {
                               "type": "boolean"
@@ -10536,7 +10605,7 @@
                       "additionalProperties": false,
                       "description": "Guard table item (QUERY/OUTPUT form — includes auto-derived object_type field)"
                     },
-                    "description": "User-submitted data matching the Guard's required fields. Relation: structure must match the Guard table's column definitions. Example: [{field:'delivery_proof', value:'Qm...'}]"
+                    "description": "User-submitted data matching the Guard's required fields. Relation: structure must match the Guard table's column definitions. Example: [{field:'delivery_proof', value:'Qm...'}] PROTOCOL SLOT 254: when the Guard table has identifier 254, submit the ORDER address bound to the operation (payment_info.for_object) — never the Progress/Machine object being advanced."
                   }
                 },
                 "required": [
@@ -10546,7 +10615,7 @@
                 "additionalProperties": false,
                 "description": "One Guard's submission data: the Guard to verify plus the user-provided data that satisfies its requirements."
               },
-              "description": "User-submitted data for each Guard. Relation: one entry per Guard in the guard array; fill submission fields and resubmit via call_with_submission. PLACEMENT: this `submission` field is at the SAME level as `data` and `env` in the operation input — NOT inside `data.data`. Example structure: {tool:'onchain_operations', data:{operation_type:'order', data:{object:'my_order', progress:{...}}}, submission:{type:'submission', guard:[...], submission:[...]}}"
+              "description": "User-submitted data for each Guard. Relation: one entry per Guard in the guard array; fill submission fields and resubmit via call_with_submission. SLOT COMPLETENESS (T8 az3-mn-261013): the submission array must cover EVERY b_submission slot in the Guard's table, not just one — a Guard whose table has multiple b_submission entries (e.g. slot 0 + slot 254) aborts on-chain with 'Submission ID not found' (abort 11) when any b_submission slot has no matching entry. Check the Guard's table (each entry with b_submission needs its identifier filled) and include ALL of them. PLACEMENT: this `submission` field is at the SAME level as `data` and `env` in the operation input — NOT inside `data.data`. Example structure: {tool:'onchain_operations', data:{operation_type:'order', data:{object:'my_order', progress:{...}}}, submission:{type:'submission', guard:[...], submission:[...]}}"
             }
           },
           "required": [
@@ -10788,7 +10857,7 @@
               ]
             },
             "guard_add": {
-              "description": "Add Guard to Reward object.",
+              "description": "Add Guard to Reward object. IDIOMS (F08 — these differ from Service allocation, follow THIS shape here): recipient is ONE OF {GuardIdentifier: <u8>} | {Entity: {name_or_address: 'mark'}} | {Signer: 'signer'} (string const INSIDE the key, NOT a bare string); amount is {type: 'Fixed', value: '0.1WOW'} — value accepts a display string with token symbol ('0.1WOW') or a smallest-unit integer ('100000000'), NOT the plain balance string used by allocation `sharing`. FULL EXAMPLE: [{guard: '0xabc...', recipient: {Signer: 'signer'}, amount: {type: 'Fixed', value: '0.1WOW'}}].",
               "type": "array",
               "items": {
                 "type": "object",
@@ -10958,6 +11027,7 @@
               "description": "Unwrap CoinWrapper objects and other objects received by this Reward object and send them to the owner of its Permission object.\n\nACCEPTED FORMATS (F-06 unified receive operation block):\n• 'recently' (string literal) — auto-query and receive ALL recently received objects.\n  Use this for the common case: \"withdraw everything the object has received.\"\n  Example: receive: 'recently'\n• ReceivedNormal[] (array) — explicit list of received objects to unwrap.\n  Use this when you want to receive specific objects only (not all).\n  Example: receive: [{id: '0xobj1', type: '0x2::coin::Coin<0x2::wow::WOW>'}]\n• ReceivedBalance ({token_type, balance, received: [{id, balance, payment}]}) —\n  receive a balance record from a specific Payment/payer.\n  Use this for precise balance targeting (advanced — usually after querying\n  the object's received history via query_received).\n  Example: receive: {token_type: '0x2::wow::WOW', balance: 1000000, received: [{id: '0xrec1', balance: 1000000, payment: '0xpay1'}]}\nANTI-PATTERN: do NOT wrap in {result: ...} — pass the value directly.",
               "anyOf": [
                 {
+                  "minItems": 1,
                   "type": "array",
                   "items": {
                     "type": "object",
@@ -11089,7 +11159,7 @@
             "object"
           ],
           "additionalProperties": false,
-          "description": "On-chain Reward operations. USAGE: (1) CREATE NEW: Set 'object' field with OBJECT format {name, type_parameter, permission, ...} to create a Reward. NOTE: 'name' goes INSIDE 'object', NOT at the data root level.'permission' can be a new Permission object or reference an existing one - check 'object' field description for details. (2) OPERATE EXISTING: Set 'object' field with STRING format (object ID or name). The 'object' field is CRITICAL and REQUIRED in both cases. STRING for existing, OBJECT for new creation."
+          "description": "On-chain Reward operations. USAGE: (1) CREATE NEW: Set 'object' field with OBJECT format {name, type_parameter, permission, ...} to create a Reward. NOTE: 'name' goes INSIDE 'object', NOT at the data root level.'permission' can be a new Permission object or reference an existing one - check 'object' field description for details. (2) OPERATE EXISTING: Set 'object' field with STRING format (object ID or name). The 'object' field is CRITICAL and REQUIRED in both cases. STRING for existing, OBJECT for new creation. NAME RESOLUTION SCOPE (T8 az3-mn-261013): a LocalMark name resolves ONLY within the SIGNING account's own local mark store — when the Reward was named by a DIFFERENT account (e.g. the merchant's mark) and you operate it from another account, the name lookup fails with 'not found locally'; pass the on-chain 0x object ID instead (resolve it once via workspace/onchain_objects and reuse the id)."
         },
         "env": {
           "type": "object",
@@ -11236,6 +11306,7 @@
                           "description": "Whether this table item's value is submitted dynamically at Guard trigger time (alloc_by_guard call). \n\ntrue = value is submitted by the caller when triggering the Guard. Use for runtime-context-dependent values like order address, user address. The 'value' field is ignored when b_submission=true; the caller must provide it via submissions[]. \n\nfalse = value is static, set at Guard creation time. Use for values known when the Guard is created: expected node names, expected merchant address, expected service address. The 'value' field must be populated and will be stored on-chain permanently. \n\nRule of thumb: if the value is the SAME for all future Guard triggers, use false. If the value DIFFERS per trigger (e.g., which order to release funds for), use true."
                         },
                         "value_type": {
+                          "description": "Type of the value stored in `value`. One of: Bool(0), Address(1), String(2), U8(3), U16(4), U32(5), U64(6), U128(7), U256(8), VecBool(9), VecAddress(10), VecString(11), VecU8(12), VecU16(13), VecU32(14), VecU64(15), VecU128(16), VecU256(17), VecVecU8(18). When value_type=Address (1), the `value` field accepts a hex address string, a LocalMark name, or an AccountOrMark_Address object — see `value` field description for details.",
                           "anyOf": [
                             {
                               "type": "number",
@@ -11522,11 +11593,10 @@
                               "const": "vecvecu8",
                               "description": "vecvecu8"
                             }
-                          ],
-                          "description": "Type of the value stored in `value`. One of: Bool(0), Address(1), String(2), U8(3), U16(4), U32(5), U64(6), U128(7), U256(8), VecBool(9), VecAddress(10), VecString(11), VecU8(12), VecU16(13), VecU32(14), VecU64(15), VecU128(16), VecU256(17), VecVecU8(18). When value_type=Address (1), the `value` field accepts a hex address string, a LocalMark name, or an AccountOrMark_Address object — see `value` field description for details."
+                          ]
                         },
                         "value": {
-                          "description": "The actual value data. Format depends on `value_type`:\n• Bool: true/false (boolean)\n• Address (CRITICAL — string 'Address' is NOT a placeholder): a hex address (e.g. '0x1234...'), a LocalMark name (e.g. 'my-service' — resolved to address at evaluation time), an AccountOrMark_Address object (e.g. {name_or_address:'my-service'}), or system shorthand ('0xaaa' = EntityLinker, '0xaab' = EntityRegistrar). The literal string 'Address' itself is INVALID — it would be treated as a non-existent LocalMark name and fail. Example: value='0x2::wow::WOW<address>' or value='my-permission'.\n• String: any string\n• U8/U16/U32/U64/U128/U256: number or numeric string (e.g. 42 or '42')\n• Vec* types: arrays of the corresponding element type\nREQUIRED when b_submission=false. OPTIONAL when b_submission=true (value is supplied at evaluation time by user submission).",
+                          "description": "The actual value data. Format depends on `value_type`:\nTYPE-CLASS SPLIT (F03): scalar value_types (Bool/Address/String/U8..U256) take ONE scalar value (boolean/address string/name/string/number); Vec* value_types (VecBool/VecAddress/VecString/VecU8..VecU256) take a REAL JSON ARRAY of elements — a string (even a Python-repr-looking \"[a, b]\") fails serialization with 'value_type=10 (array) requires a array value'. The historical 'static values must be strings' rule applied to SCALARS only.\n• Bool: true/false (boolean)\n• Address (CRITICAL — string 'Address' is NOT a placeholder): a hex address (e.g. '0x1234...'), a LocalMark name (e.g. 'my-service' — resolved to address at evaluation time), an AccountOrMark_Address object (e.g. {name_or_address:'my-service'}), or system shorthand ('0xaaa' = EntityLinker, '0xaab' = EntityRegistrar). The literal string 'Address' itself is INVALID — it would be treated as a non-existent LocalMark name and fail. Example: value='0x2::wow::WOW<address>' or value='my-permission'.\n• String: any string\n• U8/U16/U32/U64/U128/U256: number or numeric string (e.g. 42 or '42')\n• Vec* types: arrays of the corresponding element type\nPROTOCOL SLOT 254 (once, b_submission Address): the submitted value MUST be the ORDER address bound to the operation (payment_info.for_object) — NEVER the Progress/Machine/Arb object the call is advancing. The chain asserts submission(254) == the order address; a Progress address fails the task binding and the forward aborts.\nREQUIRED when b_submission=false. OPTIONAL when b_submission=true (value is supplied at evaluation time by user submission).",
                           "anyOf": [
                             {
                               "type": "boolean"
@@ -11685,7 +11755,7 @@
                       "additionalProperties": false,
                       "description": "Guard table item (QUERY/OUTPUT form — includes auto-derived object_type field)"
                     },
-                    "description": "User-submitted data matching the Guard's required fields. Relation: structure must match the Guard table's column definitions. Example: [{field:'delivery_proof', value:'Qm...'}]"
+                    "description": "User-submitted data matching the Guard's required fields. Relation: structure must match the Guard table's column definitions. Example: [{field:'delivery_proof', value:'Qm...'}] PROTOCOL SLOT 254: when the Guard table has identifier 254, submit the ORDER address bound to the operation (payment_info.for_object) — never the Progress/Machine object being advanced."
                   }
                 },
                 "required": [
@@ -11695,7 +11765,7 @@
                 "additionalProperties": false,
                 "description": "One Guard's submission data: the Guard to verify plus the user-provided data that satisfies its requirements."
               },
-              "description": "User-submitted data for each Guard. Relation: one entry per Guard in the guard array; fill submission fields and resubmit via call_with_submission. PLACEMENT: this `submission` field is at the SAME level as `data` and `env` in the operation input — NOT inside `data.data`. Example structure: {tool:'onchain_operations', data:{operation_type:'order', data:{object:'my_order', progress:{...}}}, submission:{type:'submission', guard:[...], submission:[...]}}"
+              "description": "User-submitted data for each Guard. Relation: one entry per Guard in the guard array; fill submission fields and resubmit via call_with_submission. SLOT COMPLETENESS (T8 az3-mn-261013): the submission array must cover EVERY b_submission slot in the Guard's table, not just one — a Guard whose table has multiple b_submission entries (e.g. slot 0 + slot 254) aborts on-chain with 'Submission ID not found' (abort 11) when any b_submission slot has no matching entry. Check the Guard's table (each entry with b_submission needs its identifier filled) and include ALL of them. PLACEMENT: this `submission` field is at the SAME level as `data` and `env` in the operation input — NOT inside `data.data`. Example structure: {tool:'onchain_operations', data:{operation_type:'order', data:{object:'my_order', progress:{...}}}, submission:{type:'submission', guard:[...], submission:[...]}}"
             }
           },
           "required": [
@@ -12289,6 +12359,7 @@
                           "description": "Whether this table item's value is submitted dynamically at Guard trigger time (alloc_by_guard call). \n\ntrue = value is submitted by the caller when triggering the Guard. Use for runtime-context-dependent values like order address, user address. The 'value' field is ignored when b_submission=true; the caller must provide it via submissions[]. \n\nfalse = value is static, set at Guard creation time. Use for values known when the Guard is created: expected node names, expected merchant address, expected service address. The 'value' field must be populated and will be stored on-chain permanently. \n\nRule of thumb: if the value is the SAME for all future Guard triggers, use false. If the value DIFFERS per trigger (e.g., which order to release funds for), use true."
                         },
                         "value_type": {
+                          "description": "Type of the value stored in `value`. One of: Bool(0), Address(1), String(2), U8(3), U16(4), U32(5), U64(6), U128(7), U256(8), VecBool(9), VecAddress(10), VecString(11), VecU8(12), VecU16(13), VecU32(14), VecU64(15), VecU128(16), VecU256(17), VecVecU8(18). When value_type=Address (1), the `value` field accepts a hex address string, a LocalMark name, or an AccountOrMark_Address object — see `value` field description for details.",
                           "anyOf": [
                             {
                               "type": "number",
@@ -12575,11 +12646,10 @@
                               "const": "vecvecu8",
                               "description": "vecvecu8"
                             }
-                          ],
-                          "description": "Type of the value stored in `value`. One of: Bool(0), Address(1), String(2), U8(3), U16(4), U32(5), U64(6), U128(7), U256(8), VecBool(9), VecAddress(10), VecString(11), VecU8(12), VecU16(13), VecU32(14), VecU64(15), VecU128(16), VecU256(17), VecVecU8(18). When value_type=Address (1), the `value` field accepts a hex address string, a LocalMark name, or an AccountOrMark_Address object — see `value` field description for details."
+                          ]
                         },
                         "value": {
-                          "description": "The actual value data. Format depends on `value_type`:\n• Bool: true/false (boolean)\n• Address (CRITICAL — string 'Address' is NOT a placeholder): a hex address (e.g. '0x1234...'), a LocalMark name (e.g. 'my-service' — resolved to address at evaluation time), an AccountOrMark_Address object (e.g. {name_or_address:'my-service'}), or system shorthand ('0xaaa' = EntityLinker, '0xaab' = EntityRegistrar). The literal string 'Address' itself is INVALID — it would be treated as a non-existent LocalMark name and fail. Example: value='0x2::wow::WOW<address>' or value='my-permission'.\n• String: any string\n• U8/U16/U32/U64/U128/U256: number or numeric string (e.g. 42 or '42')\n• Vec* types: arrays of the corresponding element type\nREQUIRED when b_submission=false. OPTIONAL when b_submission=true (value is supplied at evaluation time by user submission).",
+                          "description": "The actual value data. Format depends on `value_type`:\nTYPE-CLASS SPLIT (F03): scalar value_types (Bool/Address/String/U8..U256) take ONE scalar value (boolean/address string/name/string/number); Vec* value_types (VecBool/VecAddress/VecString/VecU8..VecU256) take a REAL JSON ARRAY of elements — a string (even a Python-repr-looking \"[a, b]\") fails serialization with 'value_type=10 (array) requires a array value'. The historical 'static values must be strings' rule applied to SCALARS only.\n• Bool: true/false (boolean)\n• Address (CRITICAL — string 'Address' is NOT a placeholder): a hex address (e.g. '0x1234...'), a LocalMark name (e.g. 'my-service' — resolved to address at evaluation time), an AccountOrMark_Address object (e.g. {name_or_address:'my-service'}), or system shorthand ('0xaaa' = EntityLinker, '0xaab' = EntityRegistrar). The literal string 'Address' itself is INVALID — it would be treated as a non-existent LocalMark name and fail. Example: value='0x2::wow::WOW<address>' or value='my-permission'.\n• String: any string\n• U8/U16/U32/U64/U128/U256: number or numeric string (e.g. 42 or '42')\n• Vec* types: arrays of the corresponding element type\nPROTOCOL SLOT 254 (once, b_submission Address): the submitted value MUST be the ORDER address bound to the operation (payment_info.for_object) — NEVER the Progress/Machine/Arb object the call is advancing. The chain asserts submission(254) == the order address; a Progress address fails the task binding and the forward aborts.\nREQUIRED when b_submission=false. OPTIONAL when b_submission=true (value is supplied at evaluation time by user submission).",
                           "anyOf": [
                             {
                               "type": "boolean"
@@ -12738,7 +12808,7 @@
                       "additionalProperties": false,
                       "description": "Guard table item (QUERY/OUTPUT form — includes auto-derived object_type field)"
                     },
-                    "description": "User-submitted data matching the Guard's required fields. Relation: structure must match the Guard table's column definitions. Example: [{field:'delivery_proof', value:'Qm...'}]"
+                    "description": "User-submitted data matching the Guard's required fields. Relation: structure must match the Guard table's column definitions. Example: [{field:'delivery_proof', value:'Qm...'}] PROTOCOL SLOT 254: when the Guard table has identifier 254, submit the ORDER address bound to the operation (payment_info.for_object) — never the Progress/Machine object being advanced."
                   }
                 },
                 "required": [
@@ -12748,7 +12818,7 @@
                 "additionalProperties": false,
                 "description": "One Guard's submission data: the Guard to verify plus the user-provided data that satisfies its requirements."
               },
-              "description": "User-submitted data for each Guard. Relation: one entry per Guard in the guard array; fill submission fields and resubmit via call_with_submission. PLACEMENT: this `submission` field is at the SAME level as `data` and `env` in the operation input — NOT inside `data.data`. Example structure: {tool:'onchain_operations', data:{operation_type:'order', data:{object:'my_order', progress:{...}}}, submission:{type:'submission', guard:[...], submission:[...]}}"
+              "description": "User-submitted data for each Guard. Relation: one entry per Guard in the guard array; fill submission fields and resubmit via call_with_submission. SLOT COMPLETENESS (T8 az3-mn-261013): the submission array must cover EVERY b_submission slot in the Guard's table, not just one — a Guard whose table has multiple b_submission entries (e.g. slot 0 + slot 254) aborts on-chain with 'Submission ID not found' (abort 11) when any b_submission slot has no matching entry. Check the Guard's table (each entry with b_submission needs its identifier filled) and include ALL of them. PLACEMENT: this `submission` field is at the SAME level as `data` and `env` in the operation input — NOT inside `data.data`. Example structure: {tool:'onchain_operations', data:{operation_type:'order', data:{object:'my_order', progress:{...}}}, submission:{type:'submission', guard:[...], submission:[...]}}"
             }
           },
           "required": [
@@ -13429,6 +13499,7 @@
               "description": "Unwrap CoinWrapper objects and other objects received by this Permission object and send them to the builder(owner).\n\nACCEPTED FORMATS (F-06 unified receive operation block):\n• 'recently' (string literal) — auto-query and receive ALL recently received objects.\n  Use this for the common case: \"withdraw everything the object has received.\"\n  Example: receive: 'recently'\n• ReceivedNormal[] (array) — explicit list of received objects to unwrap.\n  Use this when you want to receive specific objects only (not all).\n  Example: receive: [{id: '0xobj1', type: '0x2::coin::Coin<0x2::wow::WOW>'}]\n• ReceivedBalance ({token_type, balance, received: [{id, balance, payment}]}) —\n  receive a balance record from a specific Payment/payer.\n  Use this for precise balance targeting (advanced — usually after querying\n  the object's received history via query_received).\n  Example: receive: {token_type: '0x2::wow::WOW', balance: 1000000, received: [{id: '0xrec1', balance: 1000000, payment: '0xpay1'}]}\nANTI-PATTERN: do NOT wrap in {result: ...} — pass the value directly.",
               "anyOf": [
                 {
+                  "minItems": 1,
                   "type": "array",
                   "items": {
                     "type": "object",
@@ -13713,6 +13784,7 @@
                     "description": "Whether this table item's value is submitted dynamically at Guard trigger time (alloc_by_guard call). \n\ntrue = value is submitted by the caller when triggering the Guard. Use for runtime-context-dependent values like order address, user address. The 'value' field is ignored when b_submission=true; the caller must provide it via submissions[]. \n\nfalse = value is static, set at Guard creation time. Use for values known when the Guard is created: expected node names, expected merchant address, expected service address. The 'value' field must be populated and will be stored on-chain permanently. \n\nRule of thumb: if the value is the SAME for all future Guard triggers, use false. If the value DIFFERS per trigger (e.g., which order to release funds for), use true."
                   },
                   "value_type": {
+                    "description": "Type of the value stored in `value`. One of: Bool(0), Address(1), String(2), U8(3), U16(4), U32(5), U64(6), U128(7), U256(8), VecBool(9), VecAddress(10), VecString(11), VecU8(12), VecU16(13), VecU32(14), VecU64(15), VecU128(16), VecU256(17), VecVecU8(18). When value_type=Address (1), the `value` field accepts a hex address string, a LocalMark name, or an AccountOrMark_Address object — see `value` field description for details.",
                     "anyOf": [
                       {
                         "type": "number",
@@ -13999,11 +14071,10 @@
                         "const": "vecvecu8",
                         "description": "vecvecu8"
                       }
-                    ],
-                    "description": "Type of the value stored in `value`. One of: Bool(0), Address(1), String(2), U8(3), U16(4), U32(5), U64(6), U128(7), U256(8), VecBool(9), VecAddress(10), VecString(11), VecU8(12), VecU16(13), VecU32(14), VecU64(15), VecU128(16), VecU256(17), VecVecU8(18). When value_type=Address (1), the `value` field accepts a hex address string, a LocalMark name, or an AccountOrMark_Address object — see `value` field description for details."
+                    ]
                   },
                   "value": {
-                    "description": "The actual value data. Format depends on `value_type`:\n• Bool: true/false (boolean)\n• Address (CRITICAL — string 'Address' is NOT a placeholder): a hex address (e.g. '0x1234...'), a LocalMark name (e.g. 'my-service' — resolved to address at evaluation time), an AccountOrMark_Address object (e.g. {name_or_address:'my-service'}), or system shorthand ('0xaaa' = EntityLinker, '0xaab' = EntityRegistrar). The literal string 'Address' itself is INVALID — it would be treated as a non-existent LocalMark name and fail. Example: value='0x2::wow::WOW<address>' or value='my-permission'.\n• String: any string\n• U8/U16/U32/U64/U128/U256: number or numeric string (e.g. 42 or '42')\n• Vec* types: arrays of the corresponding element type\nREQUIRED when b_submission=false. OPTIONAL when b_submission=true (value is supplied at evaluation time by user submission).",
+                    "description": "The actual value data. Format depends on `value_type`:\nTYPE-CLASS SPLIT (F03): scalar value_types (Bool/Address/String/U8..U256) take ONE scalar value (boolean/address string/name/string/number); Vec* value_types (VecBool/VecAddress/VecString/VecU8..VecU256) take a REAL JSON ARRAY of elements — a string (even a Python-repr-looking \"[a, b]\") fails serialization with 'value_type=10 (array) requires a array value'. The historical 'static values must be strings' rule applied to SCALARS only.\n• Bool: true/false (boolean)\n• Address (CRITICAL — string 'Address' is NOT a placeholder): a hex address (e.g. '0x1234...'), a LocalMark name (e.g. 'my-service' — resolved to address at evaluation time), an AccountOrMark_Address object (e.g. {name_or_address:'my-service'}), or system shorthand ('0xaaa' = EntityLinker, '0xaab' = EntityRegistrar). The literal string 'Address' itself is INVALID — it would be treated as a non-existent LocalMark name and fail. Example: value='0x2::wow::WOW<address>' or value='my-permission'.\n• String: any string\n• U8/U16/U32/U64/U128/U256: number or numeric string (e.g. 42 or '42')\n• Vec* types: arrays of the corresponding element type\nPROTOCOL SLOT 254 (once, b_submission Address): the submitted value MUST be the ORDER address bound to the operation (payment_info.for_object) — NEVER the Progress/Machine/Arb object the call is advancing. The chain asserts submission(254) == the order address; a Progress address fails the task binding and the forward aborts.\nREQUIRED when b_submission=false. OPTIONAL when b_submission=true (value is supplied at evaluation time by user submission).",
                     "anyOf": [
                       {
                         "type": "boolean"
@@ -14191,7 +14262,7 @@
                                   "type": "string"
                                 }
                               ],
-                              "description": "When specified, the query retrieves data from the associated (witness) object instead of the queried object itself. This enables cross-object queries within Guard validation logic. Can be a numeric ID (100-108) or a string name (case-insensitive, with or without the 'Type' prefix). String format is recommended for readability.\n\nAvailable WitnessTypes:\n- OrderProgress / TypeOrderProgress (100): Query an Order's associated Progress object. Use case: Check if order has reached a specific workflow node (e.g., 'complete') before allowing reward claims.\n- OrderMachine / TypeOrderMachine (101): Query an Order's associated Machine (workflow definition). Use case: Verify the workflow structure or check node configurations.\n- OrderService / TypeOrderService (102): Query an Order's parent Service object. Use case: Validate the service offering details or check service-level configurations.\n- ProgressMachine / TypeProgressMachine (103): Query a Progress's associated Machine. Use case: Access workflow definition from a Progress context.\n- ArbOrder / TypeArbOrder (104): Query an Arb's associated Order. Use case: In arbitration voting guards, verify order details like payment amount or service ID.\n- ArbArbitration / TypeArbArbitration (105): Query an Arb's parent Arbitration service. Use case: Access arbitration configuration or fee settings from within an Arb case.\n- ArbProgress / TypeArbProgress (106): Query an Arb's associated Progress. Use case: Check order workflow state during arbitration validation.\n- ArbMachine / TypeArbMachine (107): Query an Arb's associated Machine. Use case: Access workflow definition for dispute context analysis.\n- ArbService / TypeArbService (108): Query an Arb's associated Service. Use case: Verify service terms or compensation fund status during arbitration.\n\nExample: To validate that an order has reached the 'complete' node, query the Order with convert_witness='OrderProgress' and check the Progress's current_node field."
+                              "description": "When specified, the query retrieves data from the associated (witness) object instead of the queried object itself. This enables cross-object queries within Guard validation logic. Can be a numeric ID (100-109) or a string name (case-insensitive, with or without the 'Type' prefix). String format is recommended for readability.\n\nAvailable WitnessTypes:\n- OrderProgress / TypeOrderProgress (100): Query an Order's associated Progress object. Use case: Check if order has reached a specific workflow node (e.g., 'complete') before allowing reward claims.\n- OrderMachine / TypeOrderMachine (101): Query an Order's associated Machine (workflow definition). Use case: Verify the workflow structure or check node configurations.\n- OrderService / TypeOrderService (102): Query an Order's parent Service object. Use case: Validate the service offering details or check service-level configurations.\n- ProgressMachine / TypeProgressMachine (103): Query a Progress's associated Machine. Use case: Access workflow definition from a Progress context.\n- ArbOrder / TypeArbOrder (104): Query an Arb's associated Order. Use case: In arbitration voting guards, verify order details like payment amount or service ID.\n- ArbArbitration / TypeArbArbitration (105): Query an Arb's parent Arbitration service. Use case: Access arbitration configuration or fee settings from within an Arb case.\n- ArbProgress / TypeArbProgress (106): Query an Arb's associated Progress. Use case: Check order workflow state during arbitration validation.\n- ArbMachine / TypeArbMachine (107): Query an Arb's associated Machine. Use case: Access workflow definition for dispute context analysis.\n- ArbService / TypeArbService (108): Query an Arb's associated Service. Use case: Verify service terms or compensation fund status during arbitration.\n- OrderAllocation / TypeOrderAllocation (109): Query an Order's escrow Allocation. Use case: Fund gates — combine with the allocation.balance query (1051) to require balance >= a threshold before a forward advances (e.g., renewal loops).\n\nExample: To validate that an order has reached the 'complete' node, query the Order with convert_witness='OrderProgress' and check the Progress's current_node field."
                             }
                           },
                           "required": [
@@ -17047,7 +17118,7 @@
                     },
                     "file_path": {
                       "type": "string",
-                      "description": "Path to JSON or Markdown file containing Guard definition. File can define: namedNew, description, table, root, rely."
+                      "description": "ABSOLUTE path to JSON or Markdown file containing Guard definition. Field name is `file_path` (NOT `file`). File can define: namedNew, description, table, root, rely."
                     },
                     "format": {
                       "default": "json",
@@ -17619,7 +17690,7 @@
                                         "description": "vecvecu8"
                                       }
                                     ],
-                                    "description": "User available value type (number or string, e.g., 6 or 'U64' or 'u64')"
+                                    "description": "User available value type (number, numeric string, or name — e.g., 6, '6', 'U64', 'u64')"
                                   },
                                   {
                                     "type": "number",
@@ -18354,7 +18425,7 @@
                 "receive"
               ],
               "additionalProperties": false,
-              "description": "Receive mode: unwrap a CoinWrapper to the caller's wallet. Use after Allocation's alloc_by_guard creates a Payment with your address as a revenue recipient. The CoinWrapper holds your share — call this to convert it to actual coins in your wallet. Provide 'object' to unwrap a specific CoinWrapper, or OMIT it to auto-unwrap every CoinWrapper owned by the caller. AUTO-CLAIM: when the allocation SIGNER and the recipient are the SAME account, the share is credited directly during settlement and NO CoinWrapper is created — a 'no CoinWrapper found' result in that case means the funds already arrived (verify via account_balance), it is NOT a failure."
+              "description": "Receive mode: unwrap a CoinWrapper to the caller's wallet. Use after Allocation's alloc_by_guard creates a Payment with your address as a revenue recipient. The CoinWrapper holds your share — call this to convert it to actual coins in your wallet. Provide 'object' to unwrap a specific CoinWrapper, or OMIT it to auto-unwrap every CoinWrapper owned by the caller. CHAIN TRUTH: allocation settlement ALWAYS creates one CoinWrapper per recipient row on-chain — there is NO chain-level auto-claim, even when the signer IS the recipient (ptrn-ln-261010 C-04). After an alloc where you are a recipient, expect a CoinWrapper under your address and claim it here; a 'no CoinWrapper found' result means the funds were already claimed (verify via account_balance) or your share is held by another object (Order/Entity) and must be claimed through that object's own receive entry."
             }
           ],
           "description": "On-chain Payment operations. TWO modes:\n(1) CREATE: Set 'object' with {name, type_parameter, ...}, 'revenue', and 'info' to create a new Payment.\n(2) RECEIVE: Set {receive: true} to unwrap CoinWrappers to your wallet. Provide 'object' (<coinwrapper_id_or_name>) to unwrap a specific one, or OMIT it to auto-unwrap every CoinWrapper owned by the caller. type_parameter optional — auto-derived from the CoinWrapper's type."
@@ -18814,6 +18885,7 @@
               "description": "Unwrap CoinWrapper objects and other objects received by this Demand object and send them to the owner of its Permission object.\n\nACCEPTED FORMATS (F-06 unified receive operation block):\n• 'recently' (string literal) — auto-query and receive ALL recently received objects.\n  Use this for the common case: \"withdraw everything the object has received.\"\n  Example: receive: 'recently'\n• ReceivedNormal[] (array) — explicit list of received objects to unwrap.\n  Use this when you want to receive specific objects only (not all).\n  Example: receive: [{id: '0xobj1', type: '0x2::coin::Coin<0x2::wow::WOW>'}]\n• ReceivedBalance ({token_type, balance, received: [{id, balance, payment}]}) —\n  receive a balance record from a specific Payment/payer.\n  Use this for precise balance targeting (advanced — usually after querying\n  the object's received history via query_received).\n  Example: receive: {token_type: '0x2::wow::WOW', balance: 1000000, received: [{id: '0xrec1', balance: 1000000, payment: '0xpay1'}]}\nANTI-PATTERN: do NOT wrap in {result: ...} — pass the value directly.",
               "anyOf": [
                 {
+                  "minItems": 1,
                   "type": "array",
                   "items": {
                     "type": "object",
@@ -19092,6 +19164,7 @@
                           "description": "Whether this table item's value is submitted dynamically at Guard trigger time (alloc_by_guard call). \n\ntrue = value is submitted by the caller when triggering the Guard. Use for runtime-context-dependent values like order address, user address. The 'value' field is ignored when b_submission=true; the caller must provide it via submissions[]. \n\nfalse = value is static, set at Guard creation time. Use for values known when the Guard is created: expected node names, expected merchant address, expected service address. The 'value' field must be populated and will be stored on-chain permanently. \n\nRule of thumb: if the value is the SAME for all future Guard triggers, use false. If the value DIFFERS per trigger (e.g., which order to release funds for), use true."
                         },
                         "value_type": {
+                          "description": "Type of the value stored in `value`. One of: Bool(0), Address(1), String(2), U8(3), U16(4), U32(5), U64(6), U128(7), U256(8), VecBool(9), VecAddress(10), VecString(11), VecU8(12), VecU16(13), VecU32(14), VecU64(15), VecU128(16), VecU256(17), VecVecU8(18). When value_type=Address (1), the `value` field accepts a hex address string, a LocalMark name, or an AccountOrMark_Address object — see `value` field description for details.",
                           "anyOf": [
                             {
                               "type": "number",
@@ -19378,11 +19451,10 @@
                               "const": "vecvecu8",
                               "description": "vecvecu8"
                             }
-                          ],
-                          "description": "Type of the value stored in `value`. One of: Bool(0), Address(1), String(2), U8(3), U16(4), U32(5), U64(6), U128(7), U256(8), VecBool(9), VecAddress(10), VecString(11), VecU8(12), VecU16(13), VecU32(14), VecU64(15), VecU128(16), VecU256(17), VecVecU8(18). When value_type=Address (1), the `value` field accepts a hex address string, a LocalMark name, or an AccountOrMark_Address object — see `value` field description for details."
+                          ]
                         },
                         "value": {
-                          "description": "The actual value data. Format depends on `value_type`:\n• Bool: true/false (boolean)\n• Address (CRITICAL — string 'Address' is NOT a placeholder): a hex address (e.g. '0x1234...'), a LocalMark name (e.g. 'my-service' — resolved to address at evaluation time), an AccountOrMark_Address object (e.g. {name_or_address:'my-service'}), or system shorthand ('0xaaa' = EntityLinker, '0xaab' = EntityRegistrar). The literal string 'Address' itself is INVALID — it would be treated as a non-existent LocalMark name and fail. Example: value='0x2::wow::WOW<address>' or value='my-permission'.\n• String: any string\n• U8/U16/U32/U64/U128/U256: number or numeric string (e.g. 42 or '42')\n• Vec* types: arrays of the corresponding element type\nREQUIRED when b_submission=false. OPTIONAL when b_submission=true (value is supplied at evaluation time by user submission).",
+                          "description": "The actual value data. Format depends on `value_type`:\nTYPE-CLASS SPLIT (F03): scalar value_types (Bool/Address/String/U8..U256) take ONE scalar value (boolean/address string/name/string/number); Vec* value_types (VecBool/VecAddress/VecString/VecU8..VecU256) take a REAL JSON ARRAY of elements — a string (even a Python-repr-looking \"[a, b]\") fails serialization with 'value_type=10 (array) requires a array value'. The historical 'static values must be strings' rule applied to SCALARS only.\n• Bool: true/false (boolean)\n• Address (CRITICAL — string 'Address' is NOT a placeholder): a hex address (e.g. '0x1234...'), a LocalMark name (e.g. 'my-service' — resolved to address at evaluation time), an AccountOrMark_Address object (e.g. {name_or_address:'my-service'}), or system shorthand ('0xaaa' = EntityLinker, '0xaab' = EntityRegistrar). The literal string 'Address' itself is INVALID — it would be treated as a non-existent LocalMark name and fail. Example: value='0x2::wow::WOW<address>' or value='my-permission'.\n• String: any string\n• U8/U16/U32/U64/U128/U256: number or numeric string (e.g. 42 or '42')\n• Vec* types: arrays of the corresponding element type\nPROTOCOL SLOT 254 (once, b_submission Address): the submitted value MUST be the ORDER address bound to the operation (payment_info.for_object) — NEVER the Progress/Machine/Arb object the call is advancing. The chain asserts submission(254) == the order address; a Progress address fails the task binding and the forward aborts.\nREQUIRED when b_submission=false. OPTIONAL when b_submission=true (value is supplied at evaluation time by user submission).",
                           "anyOf": [
                             {
                               "type": "boolean"
@@ -19541,7 +19613,7 @@
                       "additionalProperties": false,
                       "description": "Guard table item (QUERY/OUTPUT form — includes auto-derived object_type field)"
                     },
-                    "description": "User-submitted data matching the Guard's required fields. Relation: structure must match the Guard table's column definitions. Example: [{field:'delivery_proof', value:'Qm...'}]"
+                    "description": "User-submitted data matching the Guard's required fields. Relation: structure must match the Guard table's column definitions. Example: [{field:'delivery_proof', value:'Qm...'}] PROTOCOL SLOT 254: when the Guard table has identifier 254, submit the ORDER address bound to the operation (payment_info.for_object) — never the Progress/Machine object being advanced."
                   }
                 },
                 "required": [
@@ -19551,7 +19623,7 @@
                 "additionalProperties": false,
                 "description": "One Guard's submission data: the Guard to verify plus the user-provided data that satisfies its requirements."
               },
-              "description": "User-submitted data for each Guard. Relation: one entry per Guard in the guard array; fill submission fields and resubmit via call_with_submission. PLACEMENT: this `submission` field is at the SAME level as `data` and `env` in the operation input — NOT inside `data.data`. Example structure: {tool:'onchain_operations', data:{operation_type:'order', data:{object:'my_order', progress:{...}}}, submission:{type:'submission', guard:[...], submission:[...]}}"
+              "description": "User-submitted data for each Guard. Relation: one entry per Guard in the guard array; fill submission fields and resubmit via call_with_submission. SLOT COMPLETENESS (T8 az3-mn-261013): the submission array must cover EVERY b_submission slot in the Guard's table, not just one — a Guard whose table has multiple b_submission entries (e.g. slot 0 + slot 254) aborts on-chain with 'Submission ID not found' (abort 11) when any b_submission slot has no matching entry. Check the Guard's table (each entry with b_submission needs its identifier filled) and include ALL of them. PLACEMENT: this `submission` field is at the SAME level as `data` and `env` in the operation input — NOT inside `data.data`. Example structure: {tool:'onchain_operations', data:{operation_type:'order', data:{object:'my_order', progress:{...}}}, submission:{type:'submission', guard:[...], submission:[...]}}"
             }
           },
           "required": [
@@ -19619,7 +19691,7 @@
               "additionalProperties": false
             },
             "required_info": {
-              "description": "Credential recorded on the order: the merchant's IM user address resolved from the Service's Contact (zero cost) or a WTS Proof object id (on-chain delivery proof embedding that IM address and timestamp; costs gas). The information itself is delivered via Wowok Messenger.",
+              "description": "Credential recorded on the order: the merchant's IM user address resolved from the Service's Contact (zero cost) or a WTS Proof object id (on-chain delivery proof embedding that IM address and timestamp; costs gas). The information itself is delivered via Wowok Messenger. WRITE PERMISSION: OWNER-ONLY — only the order owner (the buyer) can set or clear this field; any other signer (including the merchant/service owner) aborts on-chain (MoveAbort 1, 'Not the owner'). The merchant's role is to READ it and to receive the info over Messenger. ATTESTATION SEMANTICS: the recorded VALUE is self-attested and has no on-chain readback — never trust the value itself as a gate condition. A PRESENCE gate IS legitimate on-chain (query order.required_info is some — see pattern.required_info_evidence_gate): it proves the buyer recorded a credential before handover, without asserting what the credential says.",
               "anyOf": [
                 {
                   "type": "string",
@@ -19665,7 +19737,7 @@
                   ]
                 },
                 "message": {
-                  "description": "Operation result message.",
+                  "description": "Message attached to the forward advance; recorded permanently in on-chain History. REVIEW-TYPE FORWARDS: de-facto REQUIRED here — the review text is this message (the reward claim guard does not read it, but a no-message Review forward has failed repeatedly on mainnet until the message was added). Always include message: '<review text>' when advancing a Review-named forward.",
                   "type": "string"
                 }
               },
@@ -19742,6 +19814,7 @@
               "description": "Unwrap CoinWrapper objects or other objects received by the order and transfer them to the order owner. Consistent with `owner_receive` on other objects (arbitration/contact/demand/machine/permission/repository/reward/service/treasury).\n\nACCEPTED FORMATS (F-06 unified receive operation block):\n• 'recently' (string literal) — auto-query and receive ALL recently received objects.\n  Use this for the common case: \"withdraw everything the object has received.\"\n  Example: receive: 'recently'\n• ReceivedNormal[] (array) — explicit list of received objects to unwrap.\n  Use this when you want to receive specific objects only (not all).\n  Example: receive: [{id: '0xobj1', type: '0x2::coin::Coin<0x2::wow::WOW>'}]\n• ReceivedBalance ({token_type, balance, received: [{id, balance, payment}]}) —\n  receive a balance record from a specific Payment/payer.\n  Use this for precise balance targeting (advanced — usually after querying\n  the object's received history via query_received).\n  Example: receive: {token_type: '0x2::wow::WOW', balance: 1000000, received: [{id: '0xrec1', balance: 1000000, payment: '0xpay1'}]}\nANTI-PATTERN: do NOT wrap in {result: ...} — pass the value directly.",
               "anyOf": [
                 {
+                  "minItems": 1,
                   "type": "array",
                   "items": {
                     "type": "object",
@@ -20155,6 +20228,7 @@
                           "description": "Whether this table item's value is submitted dynamically at Guard trigger time (alloc_by_guard call). \n\ntrue = value is submitted by the caller when triggering the Guard. Use for runtime-context-dependent values like order address, user address. The 'value' field is ignored when b_submission=true; the caller must provide it via submissions[]. \n\nfalse = value is static, set at Guard creation time. Use for values known when the Guard is created: expected node names, expected merchant address, expected service address. The 'value' field must be populated and will be stored on-chain permanently. \n\nRule of thumb: if the value is the SAME for all future Guard triggers, use false. If the value DIFFERS per trigger (e.g., which order to release funds for), use true."
                         },
                         "value_type": {
+                          "description": "Type of the value stored in `value`. One of: Bool(0), Address(1), String(2), U8(3), U16(4), U32(5), U64(6), U128(7), U256(8), VecBool(9), VecAddress(10), VecString(11), VecU8(12), VecU16(13), VecU32(14), VecU64(15), VecU128(16), VecU256(17), VecVecU8(18). When value_type=Address (1), the `value` field accepts a hex address string, a LocalMark name, or an AccountOrMark_Address object — see `value` field description for details.",
                           "anyOf": [
                             {
                               "type": "number",
@@ -20441,11 +20515,10 @@
                               "const": "vecvecu8",
                               "description": "vecvecu8"
                             }
-                          ],
-                          "description": "Type of the value stored in `value`. One of: Bool(0), Address(1), String(2), U8(3), U16(4), U32(5), U64(6), U128(7), U256(8), VecBool(9), VecAddress(10), VecString(11), VecU8(12), VecU16(13), VecU32(14), VecU64(15), VecU128(16), VecU256(17), VecVecU8(18). When value_type=Address (1), the `value` field accepts a hex address string, a LocalMark name, or an AccountOrMark_Address object — see `value` field description for details."
+                          ]
                         },
                         "value": {
-                          "description": "The actual value data. Format depends on `value_type`:\n• Bool: true/false (boolean)\n• Address (CRITICAL — string 'Address' is NOT a placeholder): a hex address (e.g. '0x1234...'), a LocalMark name (e.g. 'my-service' — resolved to address at evaluation time), an AccountOrMark_Address object (e.g. {name_or_address:'my-service'}), or system shorthand ('0xaaa' = EntityLinker, '0xaab' = EntityRegistrar). The literal string 'Address' itself is INVALID — it would be treated as a non-existent LocalMark name and fail. Example: value='0x2::wow::WOW<address>' or value='my-permission'.\n• String: any string\n• U8/U16/U32/U64/U128/U256: number or numeric string (e.g. 42 or '42')\n• Vec* types: arrays of the corresponding element type\nREQUIRED when b_submission=false. OPTIONAL when b_submission=true (value is supplied at evaluation time by user submission).",
+                          "description": "The actual value data. Format depends on `value_type`:\nTYPE-CLASS SPLIT (F03): scalar value_types (Bool/Address/String/U8..U256) take ONE scalar value (boolean/address string/name/string/number); Vec* value_types (VecBool/VecAddress/VecString/VecU8..VecU256) take a REAL JSON ARRAY of elements — a string (even a Python-repr-looking \"[a, b]\") fails serialization with 'value_type=10 (array) requires a array value'. The historical 'static values must be strings' rule applied to SCALARS only.\n• Bool: true/false (boolean)\n• Address (CRITICAL — string 'Address' is NOT a placeholder): a hex address (e.g. '0x1234...'), a LocalMark name (e.g. 'my-service' — resolved to address at evaluation time), an AccountOrMark_Address object (e.g. {name_or_address:'my-service'}), or system shorthand ('0xaaa' = EntityLinker, '0xaab' = EntityRegistrar). The literal string 'Address' itself is INVALID — it would be treated as a non-existent LocalMark name and fail. Example: value='0x2::wow::WOW<address>' or value='my-permission'.\n• String: any string\n• U8/U16/U32/U64/U128/U256: number or numeric string (e.g. 42 or '42')\n• Vec* types: arrays of the corresponding element type\nPROTOCOL SLOT 254 (once, b_submission Address): the submitted value MUST be the ORDER address bound to the operation (payment_info.for_object) — NEVER the Progress/Machine/Arb object the call is advancing. The chain asserts submission(254) == the order address; a Progress address fails the task binding and the forward aborts.\nREQUIRED when b_submission=false. OPTIONAL when b_submission=true (value is supplied at evaluation time by user submission).",
                           "anyOf": [
                             {
                               "type": "boolean"
@@ -20604,7 +20677,7 @@
                       "additionalProperties": false,
                       "description": "Guard table item (QUERY/OUTPUT form — includes auto-derived object_type field)"
                     },
-                    "description": "User-submitted data matching the Guard's required fields. Relation: structure must match the Guard table's column definitions. Example: [{field:'delivery_proof', value:'Qm...'}]"
+                    "description": "User-submitted data matching the Guard's required fields. Relation: structure must match the Guard table's column definitions. Example: [{field:'delivery_proof', value:'Qm...'}] PROTOCOL SLOT 254: when the Guard table has identifier 254, submit the ORDER address bound to the operation (payment_info.for_object) — never the Progress/Machine object being advanced."
                   }
                 },
                 "required": [
@@ -20614,7 +20687,7 @@
                 "additionalProperties": false,
                 "description": "One Guard's submission data: the Guard to verify plus the user-provided data that satisfies its requirements."
               },
-              "description": "User-submitted data for each Guard. Relation: one entry per Guard in the guard array; fill submission fields and resubmit via call_with_submission. PLACEMENT: this `submission` field is at the SAME level as `data` and `env` in the operation input — NOT inside `data.data`. Example structure: {tool:'onchain_operations', data:{operation_type:'order', data:{object:'my_order', progress:{...}}}, submission:{type:'submission', guard:[...], submission:[...]}}"
+              "description": "User-submitted data for each Guard. Relation: one entry per Guard in the guard array; fill submission fields and resubmit via call_with_submission. SLOT COMPLETENESS (T8 az3-mn-261013): the submission array must cover EVERY b_submission slot in the Guard's table, not just one — a Guard whose table has multiple b_submission entries (e.g. slot 0 + slot 254) aborts on-chain with 'Submission ID not found' (abort 11) when any b_submission slot has no matching entry. Check the Guard's table (each entry with b_submission needs its identifier filled) and include ALL of them. PLACEMENT: this `submission` field is at the SAME level as `data` and `env` in the operation input — NOT inside `data.data`. Example structure: {tool:'onchain_operations', data:{operation_type:'order', data:{object:'my_order', progress:{...}}}, submission:{type:'submission', guard:[...], submission:[...]}}"
             }
           },
           "required": [
@@ -20711,6 +20784,7 @@
                           "description": "Whether this table item's value is submitted dynamically at Guard trigger time (alloc_by_guard call). \n\ntrue = value is submitted by the caller when triggering the Guard. Use for runtime-context-dependent values like order address, user address. The 'value' field is ignored when b_submission=true; the caller must provide it via submissions[]. \n\nfalse = value is static, set at Guard creation time. Use for values known when the Guard is created: expected node names, expected merchant address, expected service address. The 'value' field must be populated and will be stored on-chain permanently. \n\nRule of thumb: if the value is the SAME for all future Guard triggers, use false. If the value DIFFERS per trigger (e.g., which order to release funds for), use true."
                         },
                         "value_type": {
+                          "description": "Type of the value stored in `value`. One of: Bool(0), Address(1), String(2), U8(3), U16(4), U32(5), U64(6), U128(7), U256(8), VecBool(9), VecAddress(10), VecString(11), VecU8(12), VecU16(13), VecU32(14), VecU64(15), VecU128(16), VecU256(17), VecVecU8(18). When value_type=Address (1), the `value` field accepts a hex address string, a LocalMark name, or an AccountOrMark_Address object — see `value` field description for details.",
                           "anyOf": [
                             {
                               "type": "number",
@@ -20997,11 +21071,10 @@
                               "const": "vecvecu8",
                               "description": "vecvecu8"
                             }
-                          ],
-                          "description": "Type of the value stored in `value`. One of: Bool(0), Address(1), String(2), U8(3), U16(4), U32(5), U64(6), U128(7), U256(8), VecBool(9), VecAddress(10), VecString(11), VecU8(12), VecU16(13), VecU32(14), VecU64(15), VecU128(16), VecU256(17), VecVecU8(18). When value_type=Address (1), the `value` field accepts a hex address string, a LocalMark name, or an AccountOrMark_Address object — see `value` field description for details."
+                          ]
                         },
                         "value": {
-                          "description": "The actual value data. Format depends on `value_type`:\n• Bool: true/false (boolean)\n• Address (CRITICAL — string 'Address' is NOT a placeholder): a hex address (e.g. '0x1234...'), a LocalMark name (e.g. 'my-service' — resolved to address at evaluation time), an AccountOrMark_Address object (e.g. {name_or_address:'my-service'}), or system shorthand ('0xaaa' = EntityLinker, '0xaab' = EntityRegistrar). The literal string 'Address' itself is INVALID — it would be treated as a non-existent LocalMark name and fail. Example: value='0x2::wow::WOW<address>' or value='my-permission'.\n• String: any string\n• U8/U16/U32/U64/U128/U256: number or numeric string (e.g. 42 or '42')\n• Vec* types: arrays of the corresponding element type\nREQUIRED when b_submission=false. OPTIONAL when b_submission=true (value is supplied at evaluation time by user submission).",
+                          "description": "The actual value data. Format depends on `value_type`:\nTYPE-CLASS SPLIT (F03): scalar value_types (Bool/Address/String/U8..U256) take ONE scalar value (boolean/address string/name/string/number); Vec* value_types (VecBool/VecAddress/VecString/VecU8..VecU256) take a REAL JSON ARRAY of elements — a string (even a Python-repr-looking \"[a, b]\") fails serialization with 'value_type=10 (array) requires a array value'. The historical 'static values must be strings' rule applied to SCALARS only.\n• Bool: true/false (boolean)\n• Address (CRITICAL — string 'Address' is NOT a placeholder): a hex address (e.g. '0x1234...'), a LocalMark name (e.g. 'my-service' — resolved to address at evaluation time), an AccountOrMark_Address object (e.g. {name_or_address:'my-service'}), or system shorthand ('0xaaa' = EntityLinker, '0xaab' = EntityRegistrar). The literal string 'Address' itself is INVALID — it would be treated as a non-existent LocalMark name and fail. Example: value='0x2::wow::WOW<address>' or value='my-permission'.\n• String: any string\n• U8/U16/U32/U64/U128/U256: number or numeric string (e.g. 42 or '42')\n• Vec* types: arrays of the corresponding element type\nPROTOCOL SLOT 254 (once, b_submission Address): the submitted value MUST be the ORDER address bound to the operation (payment_info.for_object) — NEVER the Progress/Machine/Arb object the call is advancing. The chain asserts submission(254) == the order address; a Progress address fails the task binding and the forward aborts.\nREQUIRED when b_submission=false. OPTIONAL when b_submission=true (value is supplied at evaluation time by user submission).",
                           "anyOf": [
                             {
                               "type": "boolean"
@@ -21160,7 +21233,7 @@
                       "additionalProperties": false,
                       "description": "Guard table item (QUERY/OUTPUT form — includes auto-derived object_type field)"
                     },
-                    "description": "User-submitted data matching the Guard's required fields. Relation: structure must match the Guard table's column definitions. Example: [{field:'delivery_proof', value:'Qm...'}]"
+                    "description": "User-submitted data matching the Guard's required fields. Relation: structure must match the Guard table's column definitions. Example: [{field:'delivery_proof', value:'Qm...'}] PROTOCOL SLOT 254: when the Guard table has identifier 254, submit the ORDER address bound to the operation (payment_info.for_object) — never the Progress/Machine object being advanced."
                   }
                 },
                 "required": [
@@ -21170,7 +21243,7 @@
                 "additionalProperties": false,
                 "description": "One Guard's submission data: the Guard to verify plus the user-provided data that satisfies its requirements."
               },
-              "description": "User-submitted data for each Guard. Relation: one entry per Guard in the guard array; fill submission fields and resubmit via call_with_submission. PLACEMENT: this `submission` field is at the SAME level as `data` and `env` in the operation input — NOT inside `data.data`. Example structure: {tool:'onchain_operations', data:{operation_type:'order', data:{object:'my_order', progress:{...}}}, submission:{type:'submission', guard:[...], submission:[...]}}"
+              "description": "User-submitted data for each Guard. Relation: one entry per Guard in the guard array; fill submission fields and resubmit via call_with_submission. SLOT COMPLETENESS (T8 az3-mn-261013): the submission array must cover EVERY b_submission slot in the Guard's table, not just one — a Guard whose table has multiple b_submission entries (e.g. slot 0 + slot 254) aborts on-chain with 'Submission ID not found' (abort 11) when any b_submission slot has no matching entry. Check the Guard's table (each entry with b_submission needs its identifier filled) and include ALL of them. PLACEMENT: this `submission` field is at the SAME level as `data` and `env` in the operation input — NOT inside `data.data`. Example structure: {tool:'onchain_operations', data:{operation_type:'order', data:{object:'my_order', progress:{...}}}, submission:{type:'submission', guard:[...], submission:[...]}}"
             }
           },
           "required": [
@@ -21363,7 +21436,7 @@
               ]
             },
             "about_address": {
-              "description": "About address. Accepts null/undefined (→ Move Option::none). When provided, must be a raw 0x-prefixed address (NOT a name). e.g. address of the entity being proved.",
+              "description": "About binding (K3 contract): ONLY the counterparty ACCOUNT address (address first byte = account type) or a Proof object address (first byte = Proof type) — a messenger contact record. NEVER a business object (Progress/Order/Arb) and never a guard decision condition. Accepts null/undefined (→ Move Option::none). When provided, must be a raw 0x-prefixed address (NOT a name). e.g. the peer account being proved with.",
               "anyOf": [
                 {
                   "type": "string",
@@ -21530,6 +21603,7 @@
                           "description": "Whether this table item's value is submitted dynamically at Guard trigger time (alloc_by_guard call). \n\ntrue = value is submitted by the caller when triggering the Guard. Use for runtime-context-dependent values like order address, user address. The 'value' field is ignored when b_submission=true; the caller must provide it via submissions[]. \n\nfalse = value is static, set at Guard creation time. Use for values known when the Guard is created: expected node names, expected merchant address, expected service address. The 'value' field must be populated and will be stored on-chain permanently. \n\nRule of thumb: if the value is the SAME for all future Guard triggers, use false. If the value DIFFERS per trigger (e.g., which order to release funds for), use true."
                         },
                         "value_type": {
+                          "description": "Type of the value stored in `value`. One of: Bool(0), Address(1), String(2), U8(3), U16(4), U32(5), U64(6), U128(7), U256(8), VecBool(9), VecAddress(10), VecString(11), VecU8(12), VecU16(13), VecU32(14), VecU64(15), VecU128(16), VecU256(17), VecVecU8(18). When value_type=Address (1), the `value` field accepts a hex address string, a LocalMark name, or an AccountOrMark_Address object — see `value` field description for details.",
                           "anyOf": [
                             {
                               "type": "number",
@@ -21816,11 +21890,10 @@
                               "const": "vecvecu8",
                               "description": "vecvecu8"
                             }
-                          ],
-                          "description": "Type of the value stored in `value`. One of: Bool(0), Address(1), String(2), U8(3), U16(4), U32(5), U64(6), U128(7), U256(8), VecBool(9), VecAddress(10), VecString(11), VecU8(12), VecU16(13), VecU32(14), VecU64(15), VecU128(16), VecU256(17), VecVecU8(18). When value_type=Address (1), the `value` field accepts a hex address string, a LocalMark name, or an AccountOrMark_Address object — see `value` field description for details."
+                          ]
                         },
                         "value": {
-                          "description": "The actual value data. Format depends on `value_type`:\n• Bool: true/false (boolean)\n• Address (CRITICAL — string 'Address' is NOT a placeholder): a hex address (e.g. '0x1234...'), a LocalMark name (e.g. 'my-service' — resolved to address at evaluation time), an AccountOrMark_Address object (e.g. {name_or_address:'my-service'}), or system shorthand ('0xaaa' = EntityLinker, '0xaab' = EntityRegistrar). The literal string 'Address' itself is INVALID — it would be treated as a non-existent LocalMark name and fail. Example: value='0x2::wow::WOW<address>' or value='my-permission'.\n• String: any string\n• U8/U16/U32/U64/U128/U256: number or numeric string (e.g. 42 or '42')\n• Vec* types: arrays of the corresponding element type\nREQUIRED when b_submission=false. OPTIONAL when b_submission=true (value is supplied at evaluation time by user submission).",
+                          "description": "The actual value data. Format depends on `value_type`:\nTYPE-CLASS SPLIT (F03): scalar value_types (Bool/Address/String/U8..U256) take ONE scalar value (boolean/address string/name/string/number); Vec* value_types (VecBool/VecAddress/VecString/VecU8..VecU256) take a REAL JSON ARRAY of elements — a string (even a Python-repr-looking \"[a, b]\") fails serialization with 'value_type=10 (array) requires a array value'. The historical 'static values must be strings' rule applied to SCALARS only.\n• Bool: true/false (boolean)\n• Address (CRITICAL — string 'Address' is NOT a placeholder): a hex address (e.g. '0x1234...'), a LocalMark name (e.g. 'my-service' — resolved to address at evaluation time), an AccountOrMark_Address object (e.g. {name_or_address:'my-service'}), or system shorthand ('0xaaa' = EntityLinker, '0xaab' = EntityRegistrar). The literal string 'Address' itself is INVALID — it would be treated as a non-existent LocalMark name and fail. Example: value='0x2::wow::WOW<address>' or value='my-permission'.\n• String: any string\n• U8/U16/U32/U64/U128/U256: number or numeric string (e.g. 42 or '42')\n• Vec* types: arrays of the corresponding element type\nPROTOCOL SLOT 254 (once, b_submission Address): the submitted value MUST be the ORDER address bound to the operation (payment_info.for_object) — NEVER the Progress/Machine/Arb object the call is advancing. The chain asserts submission(254) == the order address; a Progress address fails the task binding and the forward aborts.\nREQUIRED when b_submission=false. OPTIONAL when b_submission=true (value is supplied at evaluation time by user submission).",
                           "anyOf": [
                             {
                               "type": "boolean"
@@ -21979,7 +22052,7 @@
                       "additionalProperties": false,
                       "description": "Guard table item (QUERY/OUTPUT form — includes auto-derived object_type field)"
                     },
-                    "description": "User-submitted data matching the Guard's required fields. Relation: structure must match the Guard table's column definitions. Example: [{field:'delivery_proof', value:'Qm...'}]"
+                    "description": "User-submitted data matching the Guard's required fields. Relation: structure must match the Guard table's column definitions. Example: [{field:'delivery_proof', value:'Qm...'}] PROTOCOL SLOT 254: when the Guard table has identifier 254, submit the ORDER address bound to the operation (payment_info.for_object) — never the Progress/Machine object being advanced."
                   }
                 },
                 "required": [
@@ -21989,7 +22062,7 @@
                 "additionalProperties": false,
                 "description": "One Guard's submission data: the Guard to verify plus the user-provided data that satisfies its requirements."
               },
-              "description": "User-submitted data for each Guard. Relation: one entry per Guard in the guard array; fill submission fields and resubmit via call_with_submission. PLACEMENT: this `submission` field is at the SAME level as `data` and `env` in the operation input — NOT inside `data.data`. Example structure: {tool:'onchain_operations', data:{operation_type:'order', data:{object:'my_order', progress:{...}}}, submission:{type:'submission', guard:[...], submission:[...]}}"
+              "description": "User-submitted data for each Guard. Relation: one entry per Guard in the guard array; fill submission fields and resubmit via call_with_submission. SLOT COMPLETENESS (T8 az3-mn-261013): the submission array must cover EVERY b_submission slot in the Guard's table, not just one — a Guard whose table has multiple b_submission entries (e.g. slot 0 + slot 254) aborts on-chain with 'Submission ID not found' (abort 11) when any b_submission slot has no matching entry. Check the Guard's table (each entry with b_submission needs its identifier filled) and include ALL of them. PLACEMENT: this `submission` field is at the SAME level as `data` and `env` in the operation input — NOT inside `data.data`. Example structure: {tool:'onchain_operations', data:{operation_type:'order', data:{object:'my_order', progress:{...}}}, submission:{type:'submission', guard:[...], submission:[...]}}"
             }
           },
           "required": [
@@ -22061,7 +22134,7 @@
           ]
         },
         "about_address": {
-          "description": "About address. Accepts null/undefined (→ Move Option::none). When provided, must be a raw 0x-prefixed address (NOT a name).",
+          "description": "About binding (K3 contract): ONLY the counterparty ACCOUNT address (address first byte = account type) or a Proof object address (first byte = Proof type) — a messenger contact record. NEVER a business object (Progress/Order/Arb) and never a guard decision condition. Accepts null/undefined (→ Move Option::none). When provided, must be a raw 0x-prefixed address (NOT a name).",
           "anyOf": [
             {
               "type": "string",
@@ -22589,7 +22662,7 @@
                       "description": "vecvecu8"
                     }
                   ],
-                  "description": "User available value type (number or string, e.g., 6 or 'U64' or 'u64')"
+                  "description": "User available value type (number, numeric string, or name — e.g., 6, '6', 'U64', 'u64')"
                 },
                 {
                   "type": "number",
@@ -22693,7 +22766,7 @@
                       "type": "string"
                     }
                   ],
-                  "description": "When specified, the query retrieves data from the associated (witness) object instead of the queried object itself. This enables cross-object queries within Guard validation logic. Can be a numeric ID (100-108) or a string name (case-insensitive, with or without the 'Type' prefix). String format is recommended for readability.\n\nAvailable WitnessTypes:\n- OrderProgress / TypeOrderProgress (100): Query an Order's associated Progress object. Use case: Check if order has reached a specific workflow node (e.g., 'complete') before allowing reward claims.\n- OrderMachine / TypeOrderMachine (101): Query an Order's associated Machine (workflow definition). Use case: Verify the workflow structure or check node configurations.\n- OrderService / TypeOrderService (102): Query an Order's parent Service object. Use case: Validate the service offering details or check service-level configurations.\n- ProgressMachine / TypeProgressMachine (103): Query a Progress's associated Machine. Use case: Access workflow definition from a Progress context.\n- ArbOrder / TypeArbOrder (104): Query an Arb's associated Order. Use case: In arbitration voting guards, verify order details like payment amount or service ID.\n- ArbArbitration / TypeArbArbitration (105): Query an Arb's parent Arbitration service. Use case: Access arbitration configuration or fee settings from within an Arb case.\n- ArbProgress / TypeArbProgress (106): Query an Arb's associated Progress. Use case: Check order workflow state during arbitration validation.\n- ArbMachine / TypeArbMachine (107): Query an Arb's associated Machine. Use case: Access workflow definition for dispute context analysis.\n- ArbService / TypeArbService (108): Query an Arb's associated Service. Use case: Verify service terms or compensation fund status during arbitration.\n\nExample: To validate that an order has reached the 'complete' node, query the Order with convert_witness='OrderProgress' and check the Progress's current_node field."
+                  "description": "When specified, the query retrieves data from the associated (witness) object instead of the queried object itself. This enables cross-object queries within Guard validation logic. Can be a numeric ID (100-109) or a string name (case-insensitive, with or without the 'Type' prefix). String format is recommended for readability.\n\nAvailable WitnessTypes:\n- OrderProgress / TypeOrderProgress (100): Query an Order's associated Progress object. Use case: Check if order has reached a specific workflow node (e.g., 'complete') before allowing reward claims.\n- OrderMachine / TypeOrderMachine (101): Query an Order's associated Machine (workflow definition). Use case: Verify the workflow structure or check node configurations.\n- OrderService / TypeOrderService (102): Query an Order's parent Service object. Use case: Validate the service offering details or check service-level configurations.\n- ProgressMachine / TypeProgressMachine (103): Query a Progress's associated Machine. Use case: Access workflow definition from a Progress context.\n- ArbOrder / TypeArbOrder (104): Query an Arb's associated Order. Use case: In arbitration voting guards, verify order details like payment amount or service ID.\n- ArbArbitration / TypeArbArbitration (105): Query an Arb's parent Arbitration service. Use case: Access arbitration configuration or fee settings from within an Arb case.\n- ArbProgress / TypeArbProgress (106): Query an Arb's associated Progress. Use case: Check order workflow state during arbitration validation.\n- ArbMachine / TypeArbMachine (107): Query an Arb's associated Machine. Use case: Access workflow definition for dispute context analysis.\n- ArbService / TypeArbService (108): Query an Arb's associated Service. Use case: Verify service terms or compensation fund status during arbitration.\n- OrderAllocation / TypeOrderAllocation (109): Query an Order's escrow Allocation. Use case: Fund gates — combine with the allocation.balance query (1051) to require balance >= a threshold before a forward advances (e.g., renewal loops).\n\nExample: To validate that an order has reached the 'complete' node, query the Order with convert_witness='OrderProgress' and check the Progress's current_node field."
                 }
               },
               "required": [