@adhd/backlog 1.0.5 → 1.0.6

package/skill/SKILL.md

@@ -19,40 +19,68 @@ with a `body` mints a successor with a fresh `uid` and joins the two with a
 may no longer be the _live_ one — addressing it returns `conflict` and names
 the successor. Never treat a stored uid as immutable across edits.
 
-Every example below was run against `entrypoint/backlog/dist/index.js` — the
-baseline examples on revision `9df2a5c7`, the §8 stats/rollup examples on the
-build that first mounted those ops — and its exact output is what is shown. A
+Every example below was run against `entrypoint/backlog/dist/index.js` and its
+exact output is what is shown. The obligations/verdict (§4), spec-pointer (§5),
+and catalog/order (§6) examples were run on the build that mounts all 29 verbs;
+the §11 stats/rollup examples on the build that first mounted those ops. A
 _globally installed_ `adhd-backlog` may be an older build: in particular
-`gitContext` on `create`/`transition` (§6) exists in the `9df2a5c7` build but
-an older installed build rejects it with `invalid_argument`, and the §8 stats
-ops (`priority-matrix`/`part-of-rollup`/`open-curve`) exist only in a build at
-or after the one that mounted them. Compare the `backlog create` and `backlog
-priority-matrix` lines of `adhd-backlog --help` with §1 before relying on a
-field.
-
-## 1. Command surface — 17 verbs (plus `batch`), one calling convention
-
-**Every verb takes a single `--input` flag carrying one JSON object.** There
-are no per-field flags.
+`gitContext` on `create`/`transition` (§9) exists in the `9df2a5c7` build but
+an older installed build rejects it with `invalid_argument`, and the §11 stats
+ops (`priority-matrix`/`part-of-rollup`/`open-curve`/`report`) exist only in a build at
+or after the one that mounted them. Compare the `backlog create`, `backlog
+priority-matrix`, and `backlog obligate` lines of `adhd-backlog --help` with §1
+before relying on a field.
+
+## 1. Command surface — 29 verbs (plus `batch`), one calling convention
+
+**Every verb except `embedding-status` takes a single `--input` flag carrying
+one JSON object** (`embedding-status` takes no options at all). There
+are no per-field flags: a per-field option (`get --uid …`, `query --view
+list`, `create --title …`, `batch action --operation …`) is rejected with
+`invalid_argument` (exit 2) and the message
+`Unknown option: --<field>. Available: --input`. The generated `--help`
+footer's "per-field flags are also accepted" line is boilerplate that does
+**not** hold for these commands — `--input` is the only option any verb
+accepts, verified by running it. (The special commands `serve`,
+`install-skill`, and `search` are the exception: they take argv flags and no
+`--input`.)
+
+Any verb that takes a `uid` also accepts a **unique uid prefix** (8+ hex
+characters — the first UUID block, e.g. `4fc3704e`). An exact uid always wins;
+a prefix matching two or more live nodes is refused with `ambiguous_reference`
+(exit 1) naming every candidate, and an unmatched prefix is `item_not_found`.
+A prefix shorter than 8 characters is refused as too short.
 
 ```
+adhd-backlog backlog add-citation       --input '<IAddCitationInput json>'
+adhd-backlog backlog attest             --input '<IAttestInput json>'
+adhd-backlog backlog claim              --input '<IClaimInput json>'
+adhd-backlog backlog create             --input '<ICreateIssueInput json>'
+adhd-backlog backlog delete             --input '<IDeleteIssueInput json>'
+adhd-backlog backlog embedding-status                (no options — takes neither --input nor any flag)
 adhd-backlog backlog get                --input '<IIssueGetInput json>'
-adhd-backlog backlog query              --input '<IIssueQueryInput json>'
-adhd-backlog backlog priority-matrix    --input '<IPriorityMatrixInput json>'
-adhd-backlog backlog part-of-rollup     --input '<IPartOfRollupInput json>'
+adhd-backlog backlog lookup             --input '<ILookupInput json>'
+adhd-backlog backlog merge-project      --input '<IMergeProjectInput json>'
+adhd-backlog backlog move               --input '<IMoveIssueInput json>'
+adhd-backlog backlog obligate           --input '<IObligateInput json>'
 adhd-backlog backlog open-curve         --input '<IOpenCurveInput json>'
-adhd-backlog backlog lookup             --input '{"q": "<tool, file path, or URL>"}'
-adhd-backlog backlog create             --input '<ICreateIssueInput json>'
-adhd-backlog backlog update             --input '<IUpdateIssueInput json>'
-adhd-backlog backlog transition         --input '<ITransitionInput json>'
-adhd-backlog backlog claim              --input '<IClaimInput json>'
+adhd-backlog backlog part-of-rollup     --input '<IPartOfRollupInput json>'
+adhd-backlog backlog priority-matrix    --input '<IPriorityMatrixInput json>'
+adhd-backlog backlog query              --input '<IIssueQueryInput json>'
+adhd-backlog backlog recheck            --input '<IRecheckInput json>'
 adhd-backlog backlog relate             --input '<IRelateInput json>'
-adhd-backlog backlog move               --input '<IMoveIssueInput json>'
-adhd-backlog backlog upsert-project     --input '<IUpsertProjectInput json>'
+adhd-backlog backlog remove-citation    --input '<IRemoveCitationInput json>'
+adhd-backlog backlog report             --input '<IReportInput json>'
+adhd-backlog backlog rm-location        --input '<IRmLocationInput json>'
+adhd-backlog backlog rm-project         --input '<IRmProjectInput json>'
+adhd-backlog backlog spec-append        --input '<ISpecAppendInput json>'
+adhd-backlog backlog spec-check         --input '<ISpecCheckInput json>'
+adhd-backlog backlog transition         --input '<ITransitionInput json>'
+adhd-backlog backlog unobligate         --input '<IUnobligateInput json>'
+adhd-backlog backlog update             --input '<IUpdateIssueInput json>'
 adhd-backlog backlog upsert-component   --input '<IUpsertComponentInput json>'
 adhd-backlog backlog upsert-location    --input '<IUpsertLocationInput json>'
-adhd-backlog backlog rm-location        --input '<IRmLocationInput json>'
-adhd-backlog backlog delete             --input '<IDeleteIssueInput json>'
+adhd-backlog backlog upsert-project     --input '<IUpsertProjectInput json>'
 adhd-backlog batch action               --input '<IBatchActionInput json>'
 ```
 
@@ -67,24 +95,39 @@ accepts both `adhd-backlog get --input …` and
 $ adhd-backlog --help
 Available commands:
 
+Namespaced verbs (namespace + verb, per-field flags):
+  batch action  { input: { operation: 'backlog/get', items: object[], concurrency?: number, mode?: 'parallel'|'serial'|'chained', onItemError?: 'continue'|'abort', itemTimeoutMs?: number } }
+
+Verbs (one-token, --input JSON envelope):
+  backlog add-citation  { input: { uid: string, citation: object, by: string } }
+  backlog attest  { input: { subject: object, claim: object, anchor: object, by: string } }
   backlog claim  { input: { uid: string, by: string, action: 'claim'|'release'|'renew', force?: boolean } }
-  backlog create  { input: { title: string, body: string, project: string, component?: string, kind?: string, status?: string, priority?: string, citations?: object[], author?: string, assignee?: string, gitContext?: string, by: string, duplicateAction?: 'abort'|'force'|'comment', awaitEmbed?: boolean } }
+  backlog create  { input: { title: string, body: string, project: string, component?: string, kind?: string, status?: string, priority?: string, citations?: object[], author?: string, assignee?: string, gitContext?: string, dedupeExcludeUid?: string, by: string, duplicateAction?: 'abort'|'force'|'comment', awaitEmbed?: boolean } }
   backlog delete  { input: { uid: string, reason: string, by: string, awaitEmbed?: boolean } }
-  backlog get  { input: { uid: string, fields?: union[] } | { registry: 'project'|'component'|'location', name: string, filter?: object } }
-  backlog lookup  { input: { q: string } }
+  backlog embedding-status
+  backlog get  { input: { uid: string, fields?: union[], lastN?: number, after?: string, deriveThrough?: 1|2|3|4|5 } | { registry: 'project'|'component'|'location', name: string, filter?: object } }
+  backlog lookup  { input: { q: string, kind?: string } }
+  backlog merge-project  { input: { fromUid: string, toUid: string, by: string } }
   backlog move  { input: { uid: string, toProject?: string, toComponent?: string, by: string } }
+  backlog obligate  { input: { uid: string, applies_to: object, requirement: union, on_fail: 'block'|'warn', override?: object, by: string } }
   backlog open-curve  { input: { filter?: object, at: string[] } }
-  backlog part-of-rollup  { input: { uid: string } }
+  backlog part-of-rollup  { input: { uid: string, countOnly?: boolean, limit?: number, after?: string } }
   backlog priority-matrix  { input: { filter?: object } }
-  backlog query  { input: { text?: string, filter?: object, fields?: union[], sort?: 'priority'|'updated'|'created'|'relevance'|'textMatch', direction?: 'asc'|'desc', limit?: number, offset?: number, after?: string, view?: 'list'|'ready'|'graph'|'order'|'stale'|'similar'|'overlap'|'projects'|'components'|'locations', format?: 'json'|'markdown', overlapAxis?: 'file'|'project'|'component'|'author', overlapUids?: string[], staleAfterMin?: number } }
-  backlog relate  { input: { sourceUid: string, targetUid: string, rel: 'relates_to'|'supersedes'|'blocks'|'duplicate_of'|'part_of', action: 'add'|'remove', by: string } }
+  backlog query  { input: { text?: string, filter?: object, fields?: union[], sort?: 'priority'|'updated'|'created'|'relevance'|'textMatch', direction?: 'asc'|'desc', limit?: number, offset?: number, after?: string, view?: 'list'|'ready'|'graph'|'order'|'stale'|'similar'|'overlap'|'projects'|'components'|'locations'|'kinds'|'catalogs', format?: 'json'|'markdown', overlapAxis?: 'file'|'project'|'component'|'author', overlapUids?: string[], staleAfterMin?: number, catalog?: 'kind'|'status'|'priority'|'relation'|'field'|'error_code'|'location_type'|'verb' } }
+  backlog recheck  { input: { attestationUid: string, by: string } }
+  backlog relate  { input: { sourceUid: string, targetUid: string, rel: 'relates_to'|'supersedes'|'blocks'|'duplicate_of'|'part_of'|'similar_to', action: 'add'|'remove', by: string } }
+  backlog remove-citation  { input: { uid: string, by: string, reason?: string } }
+  backlog report  { input: { filter?: object } }
   backlog rm-location  { input: { uid: string, by: string, reason?: string } }
-  backlog transition  { input: { uid: string, by: string, toStatus: string, note?: string, citations?: object[], gitContext?: string } }
-  backlog update  { input: { uid: string, by: string, title?: string, body?: string, kind?: string, priority?: string, assignee?: string, author?: string, awaitEmbed?: boolean } }
+  backlog rm-project  { input: { uid: string, reason: string, by: string } }
+  backlog spec-append  { input: { uid: string, fragment: string, anchor?: object, base_revision: string, by: string } }
+  backlog spec-check  { input: { uid: string, token?: string } }
+  backlog transition  { input: { uid: string, by: string, toStatus: string, note?: string, citations?: object[], gitContext?: string, override?: object } }
+  backlog unobligate  { input: { obligationUid: string, by: string } }
+  backlog update  { input: { uid: string, by: string, title?: string, body?: string, kind?: string, priority?: string, assignee?: string, author?: string, citations?: object[], awaitEmbed?: boolean } }
   backlog upsert-component  { input: { project: string, name: string, path?: string, description?: string, by: string } }
   backlog upsert-location  { input: { component: string, project?: string, locType: 'path'|'url'|'tool', value: string, by: string } }
   backlog upsert-project  { input: { name: string, path?: string, repoUrl?: string, monorepo?: boolean, description?: string, by: string } }
-  batch action  { input: { operation: 'backlog/get', items: object[], concurrency?: number, mode?: 'parallel'|'serial'|'chained', onItemError?: 'continue'|'abort', itemTimeoutMs?: number } }
 ```
 
 Trust that output over anything hardcoded here — it is the live schema, not
@@ -114,6 +157,23 @@ Combined with the global `--namespace sandbox` flag (valid before any
 command) it reports the throwaway store that flag would mint, so you can
 check isolation without creating anything.
 
+The **`adhdRoot` key is present only when a root was explicitly resolved** —
+the `ADHD_ROOT` env var is set, or `--namespace sandbox` minted one — and is
+**omitted** for the default `production`/`test` namespaces. Verified: plain
+`sandbox-path` prints
+`{"namespace":"production","dbPath":"/…/backlog-v2.db","embeddingEnabled":true}`
+with no `adhdRoot`, while the same call with `ADHD_ROOT=/tmp/foreign-root` (or
+`--namespace sandbox`) adds the key.
+
+`--namespace sandbox` also **ignores a foreign `ADHD_ROOT`** rather than
+writing to it: if the variable names a path this tool did not mint as a
+sandbox, it warns
+`[backlog] --namespace sandbox: ADHD_ROOT=<path> is set but is not a sandbox this tool created — ignoring it and minting a fresh isolated store instead, so --namespace sandbox never writes into an unrecognized (possibly production) location.`
+and mints a fresh throwaway store. So a stray `ADHD_ROOT` can never redirect a
+sandboxed run into an unrecognized location — but it also means a sandbox you
+want to *reuse* must be the exact path printed when it was created, not any
+directory you set `ADHD_ROOT` to.
+
 Two conventions apply to every transcript below. **Uids are truncated with
 `…` for readability** — always pass the FULL value the previous call
 returned, never the ellipsis form. And **every invocation prints warnings on
@@ -189,13 +249,23 @@ Success is always exit `0`.
 ### MCP tool names
 
 Each verb is also an MCP tool once `.mcp.json` wires the server, named
-`backlog_<verb>` with the verb's own words snake_cased: `backlog_get`,
-`backlog_query`, `backlog_priority_matrix`, `backlog_part_of_rollup`,
-`backlog_open_curve`, `backlog_lookup`, `backlog_create`, `backlog_update`,
-`backlog_transition`, `backlog_claim`, `backlog_relate`, `backlog_move`,
-`backlog_upsert_project`, `backlog_upsert_component`,
-`backlog_upsert_location`, `backlog_rm_location`, `backlog_delete`, plus the
-un-namespaced `batch_action`.
+`backlog_<verb>` with the verb's own words snake_cased: `backlog_attest`,
+`backlog_claim`, `backlog_create`, `backlog_delete`,
+`backlog_embedding_status`,
+`backlog_get`, `backlog_lookup`, `backlog_merge_project`, `backlog_move`,
+`backlog_obligate`, `backlog_open_curve`, `backlog_part_of_rollup`,
+`backlog_priority_matrix`, `backlog_query`, `backlog_recheck`, `backlog_relate`,
+`backlog_report`, `backlog_rm_location`, `backlog_rm_project`,
+`backlog_spec_append`, `backlog_spec_check`, `backlog_transition`,
+`backlog_unobligate`, `backlog_update`, `backlog_upsert_component`,
+`backlog_upsert_location`, `backlog_upsert_project`, plus the un-namespaced
+`batch_action` — 29 verbs + `batch`.
+
+The tool list a session sees is the MCP **server process's** build, which can
+lag the CLI: a long-lived `serve` started before a verb was mounted does not
+expose it until restarted. If a verb works on the CLI but its
+`mcp__backlog__*` tool is absent, restart the server rather than assuming the
+verb is unmounted.
 
 ## 3. Issue verbs — worked examples
 
@@ -205,9 +275,9 @@ missing/blank `by` is rejected with `invalid_argument` before any write
 runs.
 
 **File a new issue.** `project` is RESOLVE-ONLY — `create` never mints one;
-register it first with `upsert-project` (§4). `component` is also resolve-only
+register it first with `upsert-project` (§7). `component` is also resolve-only
 and defaults to the project's reserved `(root)` component when omitted — pass
-it, or the item is invisible to component-scoped scans (§4, "The filing rule"):
+it, or the item is invisible to component-scoped scans (§7, "The filing rule"):
 
 ```
 $ adhd-backlog backlog create --input '{
@@ -220,8 +290,29 @@ $ adhd-backlog backlog create --input '{
 {"ok":true,"data":{"created":true,"uid":"a61ff0b6-a0f1-4189-9923-671f6cbacd4e","item":{"uid":"a61ff0b6-a0f1-4189-9923-671f6cbacd4e","title":"Flaky test in auth module","kind":"issue","status":"open","priority":"HIGH","project":"020e87f2-…","component":"dcf134ab-…","createdAt":"2026-09-17T01:22:56.056Z","author":"claude:1"}}}
 ```
 
+`data` carries three more fields beyond `created`/`uid`/`item`:
+
+- **`placementResolved`** — `'explicit'` when a caller-supplied `component`
+  resolved, or `'default-root'` when the omitted `component` fell back to the
+  project's reserved `(root)` component. A supplied component that does NOT
+  resolve still throws before this field is reached, so `'default-root'` is
+  the only signal that an item was silently filed on `(root)` (and is thus
+  invisible to component-scoped scans — §7's filing rule).
+- **`duplicateScanDegraded` / `duplicateScanDegradedReason`** — present iff
+  the pre-write dedupe scan could not run a calibrated comparison, on EVERY
+  outcome (even a successful write). The reason is a concrete
+  `'no-search-backend' | 'no-embed-query' | 'no-vector-scores'`, never a
+  generic "unavailable". Verified: with embeddings off, a `create` against a
+  non-empty store returns
+  `"duplicateScanDegraded":true,"duplicateScanDegradedReason":"no-search-backend"`.
+  Both fields are **absent on a healthy scan**. When the scan is degraded in
+  the never-resolvable `'no-embed-query'` way AND `duplicateAction` is the
+  default `'abort'`, nothing is written and the result is
+  `{created:false, reason:"duplicate-scan-degraded", duplicateScanDegraded:true,
+  duplicateScanDegradedReason:"no-embed-query"}` instead.
+
 Filing more than a handful of similar issues in a row? Use `batch action`
-(§5) instead of repeating this call.
+(§8) instead of repeating this call.
 
 `create` runs a dedupe scan (FTS + semantic, when embeddings are configured)
 BEFORE writing. `duplicateAction` (default `'abort'`) controls what happens
@@ -250,13 +341,29 @@ $ adhd-backlog backlog get --input '{"uid":"a61ff0b6-…","fields":["body","cita
 registry entry by name directly, e.g.
 `{"registry":"project","name":"demo-project"}` returns the project's
 `{uid, name, path, components, locations}` — the same data `query`'s
-`view:"projects"`/`"components"`/`"locations"` list in bulk (§4).
+`view:"projects"`/`"components"`/`"locations"` list in bulk (§7).
 
 The full field vocabulary is `uid, title, kind, status, priority, project,
-component, createdAt, updatedAt, assignee, author, closedAt` (cheap/plain)
-plus `body, citations, notes, auditTrail, blockers, related, _score,
+component, createdAt, updatedAt, assignee, author, closedAt, gitContext`
+(cheap/plain) plus `body, citations, notes, auditTrail, blockers, related,
+blocksOut, dependents, partOf, obligations, verdict, spec, similar, _score,
 _vector` (opt-in only — each costs a genuine extra read, so none is in the
-default card).
+default card). `similar` is the reviewed `similar_to` links touching an item,
+both directions (`{similarTo:[…], similarFrom:[…]}`) — a different relation
+from the reserved `duplicate_of`, so a card that does not ask for it is
+unchanged. `blocksOut` is the outbound `blocks` refs; `dependents` is the
+transitive count of nodes that reach this one via `blocks`; `partOf` is the
+single `part_of` parent (or `null`); `obligations` is the declared obligation
+views (§4); `verdict` is the derived actionability verdict (§4); `spec` is the
+spec-revision pointer (§5). `_score_kind` is **not** a requestable field — it
+is a provenance tag emitted automatically alongside `_score` whenever a score
+is requested (`'rrf'` for a fused semantic rank, `'bm25'` for a grep-only FTS
+score, `'cosine'`/`'rank'`/`'priority'` otherwise); passing it in `fields`
+rejects as an unknown field. A rank-derived score is ordinal, never a
+similarity. `get { fields:["auditTrail"], lastN:5 }` bounds the
+sub-collection to its newest tail; `after` continues from the last returned
+uid of a bounded sub-collection, and `deriveThrough:1..5` caps the verdict
+ladder rung (§4).
 
 **Search/filter/page issues:**
 
@@ -267,13 +374,27 @@ $ adhd-backlog backlog query --input '{"filter":{"project":"demo-project","statu
 
 `query.view` (default `'list'`) selects the result shape: `list` · `ready` ·
 `graph` · `order` · `stale` · `similar` · `overlap` · `projects` · `components`
-· `locations` (the last three are the registry LIST views — §4). `text` is the
+· `locations` (the last three are the registry LIST views — §7). **`ready` is
+NOT an actionability filter** — it selects the `open` issues that are
+**unclaimed** and whose every live incoming `blocks` blocker is **terminal**
+(a pure claim/`blocks` predicate; it never evaluates obligations, evidence, or
+attestations). An item whose `block`-severity obligation is unsatisfied **and
+scoped to a non-terminal transition** — `fields:["verdict"]` returns
+`actionable:false` — still appears in `ready` (a *close*-scoped obligation does
+not gate claimability and leaves the item `actionable:true`), while an item
+with an open `blocks` source does not. **To find actionable
+work, read `fields:["verdict"]`; do not send an actionable-work query to
+`view:"ready"`.** `text` is the
 natural-language form — routed to `filter.semantic` when a populated vector
 space can rank it, or `filter.grep` (keyword FTS) otherwise; never set
 `text` alongside `filter.semantic`/`filter.grep` yourself. Pagination is
 truthful: `meta.total` is the count before `limit`/`offset`, `meta.returned`
 is `data.items.length`, and a page cut short for any reason other than your
-own `limit` sets `meta.truncated`.
+own `limit` sets `meta.truncated`. The four item-list views
+(`list`/`ready`/`stale`/`similar`) also carry `meta.has_more`; where a true
+pre-limit total is unknowable at bounded cost (`ready`/`stale`/`similar`),
+`meta.total_relation: 'gte'` labels `meta.total` as a lower bound.
+`graph`/`order`/`overlap` carry no `meta` by design.
 
 **Edit an existing issue.** Every field edits in place **except `body`**: a
 `body` change SUPERSEDES the issue, minting a successor node with a fresh
@@ -299,6 +420,38 @@ $ adhd-backlog backlog get --input '{"uid":"a61ff0b6-…"}'
 {"ok":false,"error":{"code":"conflict","message":"Issue \"a61ff0b6-…\" was superseded by a body edit and is no longer the live issue; it now lives under \"e3b32183-…\"","details":{"retryable":false}}}
 ```
 
+**Add or remove a single citation.** `add-citation` attaches one citation to an
+existing issue (a `citation` node + a `has_citation` edge + an audit row, in one
+transaction) and returns the citation's own `uid`. `remove-citation` retires one
+by that `uid` (never by re-specifying `(file, lines)`) and leaves the issue's
+`uid` untouched:
+
+```
+$ adhd-backlog backlog add-citation --input '{"uid":"a61ff0b6-…","by":"claude:1","citation":{"file":"packages/auth/src/index.ts","lines":"1-1"}}'
+{"ok":true,"data":{"uid":"c1e2f3a4-…","issueUid":"a61ff0b6-…","citation":{"uid":"c1e2f3a4-…","file":"packages/auth/src/index.ts","lines":"1-1","sha":"9f2c…","at":"…"}}}
+
+$ adhd-backlog backlog remove-citation --input '{"uid":"c1e2f3a4-…","by":"claude:1","reason":"superseded by a rewrite"}'
+{"ok":true,"data":{"uid":"c1e2f3a4-…","invalidated":true}}
+```
+
+`update` carries a **desired citation set** as a diff: it adds what's new and
+removes what's gone against the live set, inside one transaction. It never mints
+a new issue `uid` — only a `body` edit supersedes:
+
+```
+$ adhd-backlog backlog update --input '{"uid":"a61ff0b6-…","by":"claude:1","citations":[{"file":"packages/auth/src/index.ts","lines":"1-1"}]}'
+{"ok":true,"data":{"uid":"a61ff0b6-…","changed":["citations"]}}
+```
+
+**Cite a branch-only file with `revision`.** A citation whose `file` exists only
+on an unmerged branch is unreadable from the working tree; pin it to the branch
+and the sha is resolved from `git show <revision>:<path>` instead:
+
+```
+$ adhd-backlog backlog add-citation --input '{"uid":"a61ff0b6-…","by":"claude:1","citation":{"file":"packages/auth/src/index.ts","revision":"feat/auth"}}'
+{"ok":true,"data":{"uid":"c1e2f3a4-…","issueUid":"a61ff0b6-…","citation":{"uid":"c1e2f3a4-…","file":"packages/auth/src/index.ts","revision":"feat/auth","sha":"9f2c…","at":"…"}}}
+```
+
 **Move an issue to a new status.** A terminal `toStatus` REQUIRES `citations`
 only when the project's policy turns citation enforcement ON — `citationRequired`
 defaults to `false`, so out of the box no citation is required. When a project
@@ -348,7 +501,7 @@ $ adhd-backlog backlog claim --input '{"uid":"a61ff0b6-…","by":"claude:1","act
 ```
 
 **Link two issues.** `rel` is one of `relates_to`, `supersedes`, `blocks`,
-`duplicate_of`, `part_of` — NOT the bare word `"related"`:
+`duplicate_of`, `part_of`, `similar_to` — NOT the bare word `"related"`:
 
 ```
 $ adhd-backlog backlog relate --input '{"sourceUid":"777c5e33-…","targetUid":"a61ff0b6-…","rel":"relates_to","action":"add","by":"claude:1"}'
@@ -360,6 +513,21 @@ found none — no edge was written and no audit row produced; never assume
 every call was a fresh write. `targetUid` may belong to a different project
 than `sourceUid`.
 
+**Similarity (advisory scan, reviewed link).** Cross-project similarity is a
+two-step, deliberately non-automatic flow:
+
+- The scan surfaces **candidates**; it never writes a link. `create` reports
+  them on `similarCandidates` (scope-controlled by the project policy's
+  `similarityScope`: `same-project` default · `multi-project` · `store-wide`),
+  and `query {view:"similar", filter:{project}}` returns a `clusters` block
+  (`candidate` = advisory scan output, `linked` = existing links).
+- A reviewed link is the **existing `relate`** verb with
+  `rel:"similar_to"` (there is **no** `link-duplicate` verb). `similar_to` is
+  `n:m` issue→issue. `duplicate_of` stays **reserved** for the reviewed
+  actual-same judgement and is never written by a scan.
+- Filter reads with `filter.similarTo` (items linked `similar_to` X) or
+  `filter.hasSimilar:true` (items with ≥1 incoming link).
+
 **Move an issue to a different project/component.** `toProject`/
 `toComponent` are RESOLVE-ONLY, never minted — register the destination
 first with `upsert-project`/`upsert-component` if it doesn't exist yet:
@@ -378,7 +546,404 @@ $ adhd-backlog backlog delete --input '{"uid":"777c5e33-…","reason":"duplicate
 {"ok":true,"data":{"uid":"777c5e33-…","invalidated":true}}
 ```
 
-## 4. Registry — project / component / location
+**Attest anchored evidence.** `attest` creates a SEPARATE `attestation` node
+keyed to an issue — it never changes the issue's `uid` or revision. The anchor
+is a `locator` (`path:<file>[:<line>]`, `url:<url>`, `query:<cql>`,
+`registry:<ref>`) plus a `digest`; a `path:` anchor is checked against the
+project's git tree by a cheap-first ladder and its `check.state` is one of
+`verified` / `stale` / `unknown` / `unverified` (always present, with a
+`reason`). `recheck` APPENDS a fresh check to the record's `checks[]` history:
+
+```
+$ adhd-backlog backlog attest --input '{"subject":{"id":"777c5e33-…","revision":0},"claim":{"kind":"published-artifact"},"anchor":{"locator":"path:dist/index.js","digest":"<sha256>"},"by":"claude:1"}'
+{"ok":true,"data":{"attestationUid":"…","subject":{"id":"777c5e33-…","revision":0},"check":{"state":"verified","method":"changed_since","checked_at":"…","checked_by":"claude:1"}}}
+$ adhd-backlog backlog recheck --input '{"attestationUid":"<attestationUid>","by":"claude:1"}'
+{"ok":true,"data":{"attestationUid":"…","checks":[{"state":"verified",…},{"state":"stale",…}]}}
+```
+
+**Prerequisite for `path:` anchors — the subject project's registered `path`
+must be a git work tree.** The anchor ladder resolves a `path:` locator by
+running git against that project path, so `check.state:"verified"` is produced
+only when the path is a real git work tree containing the file. With a
+**non-git** registered project path the real result is
+`{"state":"unknown","method":"none","reason":"no git work tree at \"<path>\""}`
+— `verified` is never produced, so a `block`-severity `evidence` obligation of
+that kind **stays unsatisfied and the close stays refused**. A `url:` anchor is
+likewise only `unverified`
+(`reason:"no mechanical checker is wired for \"url:\" anchors yet"`), and a
+`revision:` anchor (a spec-revision annotation, recorded through `attest`) is
+`unverified` too. Without a git
+work tree to verify against, satisfy such an obligation another way: register
+the project with a `path` that **is** the real git work tree, use
+`on_fail:"warn"` (a warn obligation never gates a transition), or list your
+actor in the obligation's `override.actors` and pass a recorded
+`override.reason` (§4 "Overriding a refusing obligation").
+
+## 4. Obligations, attestations & the actionability verdict
+
+An **obligation** is a typed, stored requirement on an issue, scoped to a
+transition. `obligate` declares one; the write gate REFUSES a transition that
+would violate it; `fields:["verdict"]` predicts that gate ON READ. Nothing is
+evaluated at `obligate` time — the gate and the verdict do the evaluating.
+
+### Declaring an obligation
+
+`obligate { uid, applies_to, requirement, on_fail, override?, by }`:
+
+- `applies_to.to` — REQUIRED. Scopes a TRANSITION into that status: a concrete
+  status name, or `'*'` for "any terminal transition". `applies_to.from`
+  optionally narrows the source status; absent ⇒ any from-status.
+- `requirement` — the CLOSED predicate core (six productions; no CEL, no
+  dynamic leaf, no timeout). It is validated recursively BEFORE the write lock,
+  so a malformed predicate never holds it:
+  - `{ "op": "evidence", "kind": <string>, "min"?: <int ≥1> }` — at least
+    `min ?? 1` distinct **verified** attestations of `kind` (a `check.state` of
+    `verified`).
+  - `{ "op": "blockers_terminal" }` — every incoming live `blocks` source is
+    terminal (a missing status is non-terminal, fail-closed).
+  - `{ "op": "relation", "type": <edge kind>, "direction": "in"|"out" }` — a
+    live edge of `type` touches the subject (`in` = subject is `dst`, `out` =
+    `src`).
+  - `{ "op": "all_of", "of": [ … ] }` — empty list ⇒ `true`.
+  - `{ "op": "any_of", "of": [ … ] }` — empty list ⇒ `false`.
+  - `{ "op": "not", "of": { … } }`.
+- `on_fail` — `'block'` refuses the transition; `'warn'` records the shortfall
+  but allows it. A `warn` obligation never affects `actionable`.
+- `override` — `{ "actors": [<identity>, …] }`: the identities permitted to
+  override THIS obligation. An override ALWAYS requires a recorded reason
+  (below); it is never a configurable boolean.
+
+Read the obligations back with `get … fields:["obligations"]` — the STORED
+predicate, never an evaluation:
+
+```
+$ adhd-backlog get --input '{"uid":"1d9b77e5-…","fields":["obligations"]}'
+{"ok":true,"data":{"uid":"1d9b77e5-…","obligations":[{"uid":"fbcee200-…","applies_to":{"to":"closed"},"requirement":{"op":"evidence","kind":"published-artifact","min":1},"on_fail":"block"}]}}
+```
+
+### The verdict (derived on read, never stored)
+
+`fields:["verdict"]` returns the actionability verdict:
+
+```jsonc
+{ "actionable": true | false | "unknown",  // TRI-STATE — never a bare boolean
+  "evaluated_at": "…",
+  "revision": 0,                            // the issue's content revision
+  "conditions": [                           // block-severity first, then warn
+    { "type": "Evidence", "status": "True", "severity": "block",
+      "code": "EvidenceUnverified", "subject": "<obligationUid>", "message": "…" } ] }
+```
+
+`actionable` is `false` iff at least one `block` condition is `True`;
+`"unknown"` iff a `block` condition is `Unknown` and none is `True` (the honest
+"could not decide" — **never a green light**); otherwise `true`. `status` says
+whether the NAMED condition HOLDS (`{type:"Blocked",status:"True"}` ⇒ the item
+IS blocked). `type` ∈ `Blocked | Obligation | Evidence | Claim | Reference |
+Budget`; `severity` ∈ `block | warn`; the governed `code`s are `BlockedBy`,
+`MissingObligation`, `EvidenceUnverified`, `EvidenceStale`, `ClaimStale`,
+`ReferenceUnresolved`, `Unknown`, plus a `<domain>/<Code>` extension namespace.
+An obligation-free item reports a `warn` `MissingObligation` — reported, never
+blocking.
+
+The verdict runs a cheap-first ladder with a per-read budget (`deriveThrough`
+on `get`, 1–5). `get` defaults to rung 3; the list views always run rung 2:
+
+| rung | what runs                                                    | produces                                       |
+| ---- | ------------------------------------------------------------ | ---------------------------------------------- |
+| 1    | incoming live `blocks` (indexed in-degree)                   | `Blocked`/`BlockedBy` per non-terminal blocker |
+| 2    | obligation presence + close-predictor skip + predicate evaluation + claim staleness | `Obligation`/`Evidence`/`Claim` |
+| 3    | anchor existence at HEAD                                     | `EvidenceStale` when absent                    |
+| 4    | changed-since-filing                                         | `EvidenceStale` when moved since filing        |
+| 5    | full anchor re-resolve (digest match)                        | `EvidenceStale` on digest mismatch             |
+
+Where a rung was not run the condition is `Unknown`, so the list and `get`
+paths can never disagree `true` vs `false` for the same item.
+
+### End-to-end: declare → read → refuse → attest → close
+
+Declare the obligation, then read it. An obligation is scoped by
+`applies_to.to` to the **transition** it guards — a close-scoped obligation
+gates the CLOSE, not claimability — so `get` reports the item actionable (the
+same answer `claim` gives), and the close is what gets refused:
+
+```
+$ adhd-backlog obligate --input '{"uid":"1d9b77e5-…","applies_to":{"to":"closed"},"requirement":{"op":"evidence","kind":"published-artifact","min":1},"on_fail":"block","by":"agent:worker-1"}'
+{"ok":true,"data":{"uid":"1d9b77e5-…","obligationUid":"fbcee200-…"}}
+
+$ adhd-backlog get --input '{"uid":"1d9b77e5-…","fields":["verdict","obligations"]}'
+{"ok":true,"data":{"uid":"1d9b77e5-…","obligations":[{"uid":"fbcee200-…","applies_to":{"to":"closed"},"requirement":{"op":"evidence","kind":"published-artifact","min":1},"on_fail":"block"}],"verdict":{"actionable":true,"evaluated_at":"…","revision":0,"conditions":[]}}}
+```
+
+An obligation scoped to a **non-terminal** transition (e.g.
+`applies_to:{to:"in_progress"}`) IS due at claim time and appears as a rung-2
+block condition instead.
+
+Attempt the close — the gate refuses with `precondition_failed` and NOTHING is
+written (the throw rolls the transaction back; a re-read shows the status
+unchanged):
+
+```
+$ adhd-backlog transition --input '{"uid":"1d9b77e5-…","by":"agent:worker-1","toStatus":"closed","note":"attempted close before evidence"}'
+{"ok":false,"error":{"code":"precondition_failed","message":"Transition refused: EvidenceUnverified (requires \"published-artifact\") — no verified attestation of kind \"published-artifact\" satisfies this obligation","details":{"retryable":false}}}
+```
+
+Attest the evidence. `attest` mints a SEPARATE `attestation` node — the issue's
+`uid` and revision are untouched. The anchor is `locator + digest`; a `path:`
+anchor is resolved against the subject project's registered `path` by the
+cheap-first anchor ladder, and its `check.state` is one of
+`verified | stale | unknown | unverified` with a `reason` when not verified.
+`recheck { attestationUid, by }` APPENDS a fresh check to the record's
+`checks[]` history (`attest` itself returns the single `check`):
+
+```
+$ adhd-backlog attest --input '{"subject":{"id":"1d9b77e5-…","revision":0},"claim":{"kind":"published-artifact"},"anchor":{"locator":"path:README.md","digest":"<sha256 hex of the HEAD blob>"},"by":"agent:worker-1"}'
+{"ok":true,"data":{"attestationUid":"8f3604b9-…","subject":{"id":"1d9b77e5-…","revision":0},"check":{"state":"verified","method":"changed_since","checked_at":"…","checked_by":"agent:worker-1"}}}
+```
+
+That `verified` requires the subject project's registered `path` to be a **git
+work tree** containing the file. Against a non-git project path the same call
+returns `state:"unknown"`, `method:"none"`,
+`reason:"no git work tree at \"<path>\""`, and the transition below stays
+refused — see §3's `attest` note for the remedies.
+
+Now the close succeeds:
+
+```
+$ adhd-backlog transition --input '{"uid":"1d9b77e5-…","by":"agent:worker-1","toStatus":"closed","note":"evidence published"}'
+{"ok":true,"data":{"uid":"1d9b77e5-…","fromStatus":"open","toStatus":"closed","closedAt":"…","transitionUid":"f42dd635-…"}}
+```
+
+For an obligation scoped to a **non-terminal** transition — which IS evaluated
+at rung 2 — the default `get` rung (3) finds the predicate satisfied but does
+not re-resolve the anchor's freshness, so `actionable` is honestly `"unknown"`;
+raise `deriveThrough:5` to re-resolve it to `true`:
+
+```
+$ adhd-backlog get --input '{"uid":"1d9b77e5-…","fields":["verdict"]}'
+{"ok":true,"data":{"uid":"1d9b77e5-…","verdict":{"actionable":"unknown","evaluated_at":"…","revision":0,"conditions":[{"type":"Evidence","status":"Unknown","severity":"block","code":"Unknown","subject":"fbcee200-…","message":"the \"published-artifact\" anchor presence was confirmed but freshness was not re-resolved at this rung"}]}}}
+
+$ adhd-backlog get --input '{"uid":"1d9b77e5-…","fields":["verdict"],"deriveThrough":5}'
+{"ok":true,"data":{"uid":"1d9b77e5-…","verdict":{"actionable":true,"evaluated_at":"…","revision":1,"conditions":[]}}}
+```
+
+### Overriding a refusing obligation
+
+If the obligation lists your identity in `override.actors`, the transition is
+allowed with a recorded `override.reason` (a reason is ALWAYS required):
+
+```
+$ adhd-backlog obligate --input '{"uid":"34219cf9-…","applies_to":{"to":"closed"},"requirement":{"op":"evidence","kind":"never-produced","min":1},"on_fail":"block","override":{"actors":["agent:worker-1"]},"by":"agent:worker-1"}'
+{"ok":true,"data":{"uid":"34219cf9-…","obligationUid":"7dbcc8b6-…"}}
+
+$ adhd-backlog transition --input '{"uid":"34219cf9-…","by":"agent:worker-1","toStatus":"closed","note":"override recorded","override":{"reason":"reviewed by owner"}}'
+{"ok":true,"data":{"uid":"34219cf9-…","fromStatus":"open","toStatus":"closed","closedAt":"…","transitionUid":"9a006480-…"}}
+```
+
+A non-listed actor (or a blank reason) is refused:
+
+```
+$ adhd-backlog transition --input '{"uid":"c66f7368-…","by":"agent:other","toStatus":"closed","note":"not listed","override":{"reason":"trying anyway"}}'
+{"ok":false,"error":{"code":"precondition_failed","message":"Override not permitted: actor \"agent:other\" may not override obligation \"f85373dd-…\" (the actor is not listed in override.actors, or no override reason was recorded — an override always requires a reason)","details":{"retryable":false}}}
+```
+
+### Retiring an obligation
+
+`unobligate { obligationUid, by }` soft-invalidates the obligation
+(bi-temporal, never a hard delete): it stops appearing on the card and the gate
+stops refusing.
+
+```
+$ adhd-backlog unobligate --input '{"obligationUid":"fbcee200-…","by":"agent:worker-1"}'
+{"ok":true,"data":{"obligationUid":"fbcee200-…","invalidated":true}}
+```
+
+## 5. Spec revisions — the spec pointer
+
+A work product is a REVISION of its ticket (DESIGN §12). `spec-append` mints a
+new immutable `SPEC` node holding a fragment and advances the ticket's
+`meta.spec_revision` pointer IN PLACE — the ticket's `uid` is preserved and no
+prior revision is rewritten. `spec-check` tells a reader whether the token it
+holds is still current.
+
+### `spec-append { uid, fragment, anchor?, base_revision, by }`
+
+- `uid` — the ticket; any uid on its `SUPERSEDES` chain is resolved forward.
+- `fragment` — the delta appended at this revision. The base document is the
+  base revision's own fragment; the document is the fold over the chain.
+- `anchor?` — `{ locator, digest }` for the long-form file export (a
+  `revision:<uid>` locator addresses the revision itself).
+- `base_revision` — REQUIRED CAS token: the current `spec_revision` uid the
+  caller read, or `''` when the ticket has no spec yet. A stale value is
+  refused with `precondition_failed` and NOTHING is written.
+
+Outcome: `{ uid (UNCHANGED), spec_revision (the NEW revision uid),
+spec_revision_token ('sha256:<hex>'), revision_seq }`.
+
+### `spec-check { uid, token? }`
+
+- `state` ∈ `fresh | stale | unknown`; `method` ∈
+  `token | content_hash | ancestry | none`.
+- **An ABSENT `token` is `stale`** (`method:"none"`,
+  `reason:"no-token-supplied"`) — never `fresh`. A reader that holds no token
+  must not treat the spec as current.
+- A non-matching token is `stale` (`method:"token"`, `reason:"older-token"`).
+
+### The `spec` field
+
+`get … fields:["spec"]` projects the pointer onto the card — the revision uid,
+its `'sha256:<hex>'` token, and the sequence; never the revision body. Absent
+when the item has no spec.
+
+### Worked example
+
+A ticket with no spec yet passes `base_revision:""` (the appended fragment is
+`# Design\nFirst cut of the design.`):
+
+```
+$ adhd-backlog spec-append --input '{"uid":"c79b52b0-…","fragment":"# Design\nFirst cut of the design.","base_revision":"","by":"agent:worker-1"}'
+{"ok":true,"data":{"uid":"c79b52b0-…","spec_revision":"58b5dfd8-…","spec_revision_token":"sha256:d0bcba4a…","revision_seq":1}}
+
+$ adhd-backlog get --input '{"uid":"c79b52b0-…","fields":["spec"]}'
+{"ok":true,"data":{"uid":"c79b52b0-…","spec":{"spec_revision":"58b5dfd8-…","spec_revision_token":"sha256:d0bcba4a…","revision_seq":1}}}
+```
+
+Check it, with the token and then without:
+
+```
+$ adhd-backlog spec-check --input '{"uid":"c79b52b0-…","token":"sha256:d0bcba4a…"}'
+{"ok":true,"data":{"current_revision":"58b5dfd8-…","current_token":"sha256:d0bcba4a…","state":"fresh","method":"token"}}
+
+$ adhd-backlog spec-check --input '{"uid":"c79b52b0-…"}'
+{"ok":true,"data":{"current_revision":"58b5dfd8-…","current_token":"sha256:d0bcba4a…","state":"stale","method":"none","reason":"no-token-supplied"}}
+```
+
+Appending against a stale base is refused; appending against the current
+revision advances the pointer to `revision_seq:2` while the ticket `uid` stays
+the same:
+
+```
+$ adhd-backlog spec-append --input '{"uid":"c79b52b0-…","fragment":"# Design\nSecond cut.","base_revision":"stale-rev-uid","by":"agent:worker-1"}'
+{"ok":false,"error":{"code":"precondition_failed","message":"spec revision conflict: base_revision \"stale-rev-uid\" does not match the current revision \"58b5dfd8-…\" — re-read the ticket's spec_revision and retry","details":{"retryable":false}}}
+
+$ adhd-backlog spec-append --input '{"uid":"c79b52b0-…","fragment":"# Design\nSecond cut.","base_revision":"58b5dfd8-…","by":"agent:worker-1"}'
+{"ok":true,"data":{"uid":"c79b52b0-…","spec_revision":"20efdc96-…","spec_revision_token":"sha256:149a2200…","revision_seq":2}}
+```
+
+### Annotating a revision (through `attest`)
+
+There is no dedicated `annotate` verb: a comment on a spec REVISION is a plain
+`attest` call whose `subject.id` is the revision uid, whose `subject.revision`
+is its opaque `'sha256:<hex>'` token, and whose `claim.kind` is
+`"spec-annotation"`. The `anchor` is `{ locator: "revision:<revision uid>",
+digest: <token> }`. The comment is a separate `attestation` node keyed to the
+exact revision and is never written into any body, so a comment on revision _n_
+can never drift onto _n+1_.
+
+```
+$ adhd-backlog attest --input '{"subject":{"id":"20efdc96-…","revision":"sha256:149a2200…"},"claim":{"kind":"spec-annotation","body":"Reviewed: add the failure-mode section."},"anchor":{"locator":"revision:20efdc96-…","digest":"sha256:149a2200…"},"by":"agent:worker-1"}'
+{"ok":true,"data":{"attestationUid":"0fa85138-…","subject":{"id":"20efdc96-…","revision":"sha256:149a2200…"},"check":{"state":"unverified","method":"none","checked_at":"…","checked_by":"agent:worker-1","reason":"no mechanical checker is wired for \"revision:\" anchors yet"}}}
+```
+
+No mechanical checker is wired for `revision:` anchors yet, so `check.state` is
+`unverified` — the token is RECORDED, not validated.
+
+## 6. Read views — catalogs, registry lists, and dependency order
+
+`query.view` (default `'list'`) selects the result shape: `list` · `ready` ·
+`graph` · `order` · `stale` · `similar` · `overlap` · `projects` · `components`
+· `locations` · `kinds` · `catalogs`. The registry list views
+(`projects`/`components`/`locations`) are §7. The catalog views and `order` are
+below. (`ready`'s exact predicate — `open` ∧ unclaimed ∧ every live incoming
+`blocks` blocker terminal — is defined in §3; it is **not** actionability, for
+which read `fields:["verdict"]`.)
+
+### Catalogs — the live vocabularies
+
+`view:"kinds"` returns EVERY catalog: the catalog names, a `terms[]` array, and
+a case-collision flag:
+
+```
+$ adhd-backlog query --input '{"view":"kinds"}'
+{"ok":true,"data":{"view":"kinds","catalogs":{"catalogs":["kind","status","priority","relation","field","error_code","location_type","verb"],"terms":[{"name":"issue","uid":"fcfc79f5-…","catalog":"kind","source":"store","lifecycle":"active","usageCount":6},{"name":"closed","catalog":"status","source":"reserved_terminal_status_names","lifecycle":"active"},{"name":"blocks","catalog":"relation","source":"edge_kind_table","lifecycle":"active"},{"name":"verdict","catalog":"field","source":"issue_field_union","lifecycle":"active"},{"name":"precondition_failed","catalog":"error_code","source":"error_code_union","lifecycle":"active"},{"name":"path","catalog":"location_type","source":"valid_location_types","lifecycle":"active"},{"name":"obligate","catalog":"verb","source":"mounted_verb_surface","lifecycle":"active"}, …],"hasCaseCollisions":false}}}
+```
+
+`view:"catalogs"` returns the same `terms[]` at the top level, and the
+`catalog` selector narrows to ONE vocabulary (`kind` | `status` | `priority` |
+`relation` | `field` | `error_code` | `location_type` | `verb`):
+
+```
+$ adhd-backlog query --input '{"view":"catalogs","catalog":"relation"}'
+{"ok":true,"data":{"view":"catalogs","terms":[{"name":"attests","catalog":"relation","source":"edge_kind_table","lifecycle":"active"}, …]}}
+```
+
+Each term is `{ name, uid?, catalog, source, lifecycle, replacedBy?,
+usageCount? }`. `source` names the validating source it was generated from at
+call time (`store` for `kind`/`status`/`priority` — live rows with a
+`usageCount`; `edge_kind_table`, `reserved_terminal_status_names`,
+`issue_field_union`, `error_code_union`, `valid_location_types`,
+`mounted_verb_surface` for the in-code vocabularies). A `deprecated` term
+carries `replacedBy`.
+
+> **`kind` is an OPEN, free-form vocabulary — the `kind` catalog is a usage
+> CENSUS, not an allowlist.** `create`/`update` accept ANY `kind` string with
+> no validation: it is stored verbatim and the name then appears as a `kind`
+> term with `source:"store"` and a `usageCount` (verified: creating with
+> `"kind":"totally-made-up-kind"` succeeds and the term shows up on the next
+> `view:"catalogs"`). On a **fresh store the `kind` catalog is EMPTY**
+> (`{"terms":[]}`) — it reflects what has been filed, not what is permitted.
+> The conventional names are `issue` (the default, and the read path's item
+> scope) and `plan` (a plan is an ordinary issue with `kind:"plan"`; see
+> "Dependency order" below). Unlike `create`, **`filter.kind` IS validated
+> against this live census**: filtering by a name that has never been used is
+> a `validation` error naming the existing values
+> (`existing kind values: …`), so file one item with a kind before filtering
+> by it. This mirrors §3's note that `status` is an open catalog — neither is
+> a fixed enum.
+
+> **Trap — `catalog` is silently ignored without `view:"catalogs"`.** The
+> selector is honoured ONLY on `view:"catalogs"`. `{"catalog":"status"}` with
+> no `view` falls back to the default `list` view — the ordinary issue page,
+> filtered by nothing — at exit 0, with no `terms` and no error. It is not a
+> "no matches" result, and the list contents are whatever your store holds:
+
+```
+$ adhd-backlog query --input '{"catalog":"status"}'
+{"ok":true,"data":{"view":"list","items":[{"uid":"1d9b77e5-…","title":"Rebuild backlog agent docs","kind":"issue","status":"closed","priority":"HIGH"}, …],"hasMore":false},"meta":{"total":7,"returned":7,"limit":50}}
+```
+
+> Always pass `view:"catalogs"`. (`catalog` is likewise ignored on
+> `view:"kinds"`, which always returns every catalog.)
+
+### Dependency order — a plan's members
+
+A "plan" is not a separate node kind: it is an `issue` (commonly
+`kind:"plan"`) that members attach to with
+`relate(childUid, planUid, "part_of", "add")`. Three reads answer plan
+questions:
+
+- **Members:** `filter.plan` (uid or name) returns every issue with a live
+  `part_of` edge into the plan.
+- **Dependency order:** `view:"order"` + `filter.plan` returns the members
+  topologically ordered by `blocks` — `{order:{ok:true,order:[uid,…]}}`, or
+  `{ok:false,cycle:[…]}` when the `blocks` edges form a cycle.
+- **Transitive rollup:** `part-of-rollup` (§11) counts every transitive
+  `part_of` descendant, not just the direct members.
+
+```
+$ adhd-backlog query --input '{"filter":{"plan":"850f1921-…"}}'
+{"ok":true,"data":{"view":"list","items":[{"uid":"7b21007d-…","title":"Write SKILL.md inventory","kind":"issue","status":"open"},{"uid":"d56fa59f-…","title":"Write README table","kind":"issue","status":"open"}],"hasMore":false},"meta":{"total":2,"returned":2,"limit":50}}
+
+$ adhd-backlog query --input '{"view":"order","filter":{"plan":"850f1921-…"}}'
+{"ok":true,"data":{"view":"order","order":{"ok":true,"order":["7b21007d-…","d56fa59f-…"]}}}
+
+$ adhd-backlog part-of-rollup --input '{"uid":"850f1921-…"}'
+{"ok":true,"data":{"uid":"850f1921-…","childrenTotal":2,"childrenOpen":2,"childrenClosed":0,"childrenOpenUids":["7b21007d-…","d56fa59f-…"],"hasMore":false}}
+```
+
+(In the example above `7b21007d` `blocks` `d56fa59f`, so the order is
+`[7b21007d, d56fa59f]`.)
+
+## 7. Registry — project / component / location
 
 The registry answers **"where does this live, and what do I file the bug
 against?"** in one call, before you `rg`/search for it.
@@ -423,9 +988,13 @@ A component-scoped scan is how a repo's own work is found (e.g.
 | `upsert-component` | registering/updating a path _inside_ an already-registered project               | `(project, name)`             |
 | `upsert-location`  | pointing a tool/file/URL at its owning component so `lookup` resolves it         | `(component, locType, value)` |
 | `rm-location`      | retiring a location (soft-invalidate)                                            | `uid`                         |
+| `merge-project`    | collapsing a DUPLICATE project into the canonical one (reviewed, one-way)        | `(fromUid, toUid)`            |
+| `rm-project`       | retiring a project with no survivor (reviewed, one-way)                          | `uid`                         |
 
-All four are create-or-update by that key — never a duplicate row — and all
-require `by`.
+The four `upsert*`/`rm-location` rows are create-or-update by that key — never a
+duplicate row — and all require `by`. The two project-retirement verbs are
+explicit, reviewed, one-way operations (see "Consolidating duplicate projects"
+below).
 
 **Register or update a project** (create-or-update by `name`; also mints the
 project's reserved default component `(root)` on first creation):
@@ -460,12 +1029,37 @@ $ adhd-backlog backlog lookup --input '{"q":"packages/auth/src/index.ts"}'
 {"ok":true,"data":{"project":{"uid":"020e87f2-…","name":"demo-project","path":"/tmp/demo"},"component":{"uid":"41a61c6d-…","name":"auth-service","path":"packages/auth"},"location":{"uid":"a4b0dd6b-…","locType":"path","value":"packages/auth/src/index.ts"}}}
 ```
 
-`lookup` classifies `q` automatically as a tool name, file path, or URL —
-it only resolves against LOCATIONS already registered via `upsert-location`,
-never against a bare project/component name. An unregistered value returns
-`not_found` (exit 4); a path miss falls back to a suffix/prefix scan before
-giving up, and reports a `hint` when only a partial match was found — never
-a silent empty result.
+`lookup` routes `q` by shape, in order: (1) a **uid or unique uid prefix**
+returns `{"redirect":{"verb":"get","uid":…}}`; (2) an **issue title** with
+exactly one match returns the same `get` redirect, and a title matching two or
+more issues is refused (`ambiguous_reference`); (3) a **location** is
+classified as a tool name / file path / URL and resolved as before; (4) a
+**project name or `repoUrl`** resolves to its project. Pass an optional
+`"kind":"location"|"project"|"component"|"issue"` to force one branch; the
+mounted schema now types `kind` as a plain `string` (it was a closed enum), and
+only those four names select a branch. Only an
+unregistered location is still a bare `not_found` (exit 4); a path miss falls
+back to a suffix/prefix scan before giving up, and reports a `hint` when only a
+partial match was found — never a silent empty result.
+
+**Consolidating duplicate projects** — two rows registered for the same repo
+(same `repoUrl`, different `name`) split every `owns_project` read. Merge the
+duplicate INTO the canonical survivor (one `BEGIN IMMEDIATE`; every component
+is re-pointed, the duplicate is soft-retired with a one-hop redirect so its old
+NAME still resolves, and its id is never reused):
+
+```
+$ adhd-backlog backlog merge-project --input '{"fromUid":"020e87f2-…","toUid":"41a61c6d-…","by":"claude:1"}'
+{"ok":true,"data":{"survivorUid":"41a61c6d-…","retiredUid":"020e87f2-…","movedIssues":7,"retired":true}}
+```
+
+A second identical `merge-project` is an idempotent no-op. Retire a project
+with no survivor (its old name then resolves to nothing) with:
+
+```
+$ adhd-backlog backlog rm-project --input '{"uid":"020e87f2-…","reason":"stale duplicate","by":"claude:1"}'
+{"ok":true,"data":{"uid":"020e87f2-…","retired":true}}
+```
 
 **Remove a location** (soft-invalidate by `uid`):
 
@@ -514,7 +1108,7 @@ $ adhd-backlog backlog query --input '{"view":"locations","filter":{"component":
    just a missing row (filed as 49ce83b8). Run `sandbox-path` before a write
    you care about, and file through the production CLI only.
 
-## 5. Batch — N-way fan-out over one operation
+## 8. Batch — N-way fan-out over one operation
 
 `batch action` runs the SAME operation over many items. `operation` is the
 mounted operation id, namespaced as `backlog/<verb>` (not the bare verb
@@ -538,12 +1132,11 @@ status:'rejected', reason}` — `value`/`reason` is the SAME outcome envelope
 `error.code` still applies. `mode` (`'parallel'` default · `'serial'` ·
 `'chained'`), `onItemError` (`'continue'` default · `'abort'`), and
 `concurrency`/`itemTimeoutMs` govern how the fan-out runs. The valid
-`operation` values are exactly the 17 mounted verbs above (the nine issue
-verbs, the four registry verbs, and the three stats reads of §8), each
-prefixed `backlog/` — passing a bare verb name (`"create"`) is rejected with
-`invalid_argument` naming the full list.
+`operation` values are exactly the 28 mounted verbs above, each prefixed
+`backlog/` — passing a bare verb name (`"create"`) is rejected with
+`invalid_argument` naming the full list (the live error enumerates all 28).
 
-## 6. Citations — structured, not hand-typed markdown
+## 9. Citations — structured, not hand-typed markdown
 
 A citation is `{ file, lines?, context?, symbol? }`. Pass `citations` on
 `create` or on the `transition` that moves an issue into a terminal status.
@@ -583,7 +1176,7 @@ citation lines). Omit it and nothing is stored and no output changes. A
 citation's own `context` is free-text prose and is never rendered by the
 markdown projection — it cannot carry the git context.
 
-## 7. Verify writes from a NEW process
+## 10. Verify writes from a NEW process
 
 An MCP `backlog_get` served by a long-lived `serve` process can answer out
 of that process's own in-memory/uncheckpointed state. After a write you
@@ -591,13 +1184,14 @@ care about, verify by running the `adhd-backlog` CLI in a fresh shell — a
 genuinely new process — rather than re-reading through the same live MCP
 session.
 
-## 8. Stats & rollups
+## 11. Stats & rollups
 
-Three aggregate read views are first-class mounted ops — `backlog
-priority-matrix`, `backlog part-of-rollup`, `backlog open-curve` (MCP:
-`backlog_priority_matrix` / `backlog_part_of_rollup` / `backlog_open_curve`).
-All three are read-only and take no `by`, and each returns its own shape — a
-matrix, a rollup tree, a time series — so none is a `query.view` member.
+Four aggregate read ops are first-class mounted ops — `backlog
+priority-matrix`, `backlog part-of-rollup`, `backlog open-curve`, and `backlog
+report` (MCP: `backlog_priority_matrix` / `backlog_part_of_rollup` /
+`backlog_open_curve` / `backlog_report`). All four are read-only and take no
+`by`, and each returns its own shape — a matrix, a rollup tree, a time series,
+a grouped aggregate — so none is a `query.view` member.
 
 **Priority matrix** — per-priority counts, scoped by
 `project`/`component`/`kind`/`status`. An omitted `filter.status` scopes to
@@ -612,11 +1206,13 @@ $ adhd-backlog backlog priority-matrix --input '{}'
 
 **Part-of rollup** — every TRANSITIVE `part_of` descendant of the root issue
 (not just direct children), counted once each regardless of chain depth, split
-into `childrenOpen`/`childrenClosed` (plus the open descendants' uids):
+into `childrenOpen`/`childrenClosed` (plus the open descendants' uids).
+`countOnly:true` omits `childrenOpenUids` entirely (counts only); `limit`/`after`
+page the uid list (`nextCursor`/`hasMore`; default 50, max 1000):
 
 ```
 $ adhd-backlog backlog part-of-rollup --input '{"uid":"74c22c35-…"}'
-{"ok":true,"data":{"uid":"74c22c35-…","childrenTotal":1,"childrenOpen":1,"childrenClosed":0,"childrenOpenUids":["db4587ba-…"]}}
+{"ok":true,"data":{"uid":"74c22c35-…","childrenTotal":1,"childrenOpen":1,"childrenClosed":0,"childrenOpenUids":["db4587ba-…"],"hasMore":false}}
 ```
 
 **Open curve** — for each sampled ISO-8601 instant, how many in-scope issues
@@ -628,13 +1224,24 @@ $ adhd-backlog backlog open-curve --input '{"at":["2020-01-01T00:00:00.000Z"]}'
 {"ok":true,"data":{"points":[{"at":"2020-01-01T00:00:00.000Z","existed":0,"open":0,"closed":0}]}}
 ```
 
-The same three functions are also exported from the package's query layer
-(`src/query/views/stats.ts`, re-exported by `src/query/index.ts`) for
-in-process consumers — `priorityMatrix(handle, { filter? })`,
-`partOfRollup(handle, { uid })`, `openCurve(handle, { filter?, at })`. That
+**Report** — a grouped rollup whose every number is computed from the store in
+that call: `byKind`, `byPriority` (composed from `priority-matrix`), `byStatus`,
+and `avgAgeDays` of open in-scope items, with `statusScope` naming what was
+counted (`'open'` when `filter.status` is omitted):
+
+```
+$ adhd-backlog backlog report --input '{}'
+{"ok":true,"data":{"byKind":[{"kind":"issue","count":2}],"byPriority":[{"priority":"HIGH","rank":0,"count":1}],"byStatus":[{"status":"open","terminal":false,"count":2}],"avgAgeDays":0.01,"statusScope":"open","computedAt":"…"}}
+```
+
+The same functions are also exported from the package's query layer
+(`src/query/views/stats.ts` + `src/query/views/report.ts`, re-exported by
+`src/query/index.ts`) for in-process consumers — `priorityMatrix(handle, {
+filter? })`, `partOfRollup(handle, { uid, countOnly?, limit?, after? })`,
+`openCurve(handle, { filter?, at })`, `report(handle, { filter? })`. That
 in-process surface is described in `README.md` → "Library API".
 
-## 9. Hard rule — file a feature request when the tool is the friction
+## 12. Hard rule — file a feature request when the tool is the friction
 
 **Never silently work around the tool.** If a value or grouping you were asked for had no verb to produce it — you got it by reshaping raw output yourself (a client-side group-by, join, filter, count, or field-extract) — you MUST file a feature request before you finish. Two or more such reshapes in one dispatch, even inside a single command, is already more than enough. No task scope overrides this: a read-only task, "only add links", or "do not create items" does NOT exempt you.
 Inside the adhd repo: `create` a `FEAT` on project `adhd`, component `entrypoint/backlog`, with `duplicateAction:"comment"` (attaches your reproduction when it is already filed — never force); body = the exact command, the exact output, the workaround, and the outcome you wanted.