docs(spec): replace the dead tracker citations in the manifest, dataset and permission liveness ledgers with the deciding commit, or the decision in words

[objectstack-fleet]

Part of #20234

Clause-②: no

In the manifest, dataset and permission liveness ledgers, a tracker number that answers 404 is replaced. The note now cites the commit that decided it (ruling C+D, form C), or, where the number alone carried the meaning, says what was decided (form D). Only note and _note strings change: no row's status, evidence, proof or date moves. One @objectstack/spec patch changeset, because liveness/ ships in the package.

🤖 Generated with Claude Code

https://claude.ai/code/session_01Sfe5YjBLwB9J3y8fvm2xq1

[github-actions]

📓 Docs Drift Check

⚠️ 3 changed file(s) yielded no anchor (packages/spec/liveness/dataset.json, packages/spec/liveness/manifest.json, packages/spec/liveness/permission.json), so the pages documenting them are NOT COVERED by this run — this is not a clean bill of health for those files. Nothing else in this diff resolved to a documentable surface (no symbol, route or SDK anchor derived from 1 changed package(s)).

What this run could not see

• 3 changed file(s) yielded no anchor (packages/spec/liveness/dataset.json, packages/spec/liveness/manifest.json, packages/spec/liveness/permission.json) — pages documenting those are invisible to this run
• a page that states a rule by its inputs shares no identifier with the emitter that implements the rule, so an emitter-only diff cannot list it — not on this run and not on any run. Measured on fix(driver-sql): emit varchar(maxLength) for a text field a declared index keys on #11430: content/docs/protocol/objectql/types.mdx documents the text-family column mapping by the ObjectQL type names it maps FROM (text / textarea / html) while the diff changed createColumn; it went unlisted, and it was the page that diff falsified, in four places. No shared token exists to detect this on, so a rule your change carries has to be re-read by hand in the pages that restate it.
• a key NAME is not a key, so the hand re-read the line above prescribes can land on the wrong schema. The same spelling is authorable on one governed type and a [REMOVED] tombstone on another for each of active, aria, joins, objects, template, tools and version (censused on [finding] tools is a key on BOTH AgentSchema (tombstoned, dead) and SkillSchema (live, cloud-attested), so a name-based search attributes skill examples to the agent key — it produced a false stop-the-line alarm on PR #19059 #19093 over the liveness ledger's governed types, top-level keys); nothing in a search result distinguishes the two, so a grep hit on a LIVE example reads as evidence about the DEAD key. Measured on fix(spec): the agent.tools liveness row says dead — it claimed live on a key the schema tombstoned #19059: content/docs/ai/agents.mdx was reported as contradicting the agent.tools tombstone over its tools: example at :161, which is inside the defineSkill({ block opened at :155 — the page was already correct. Settle ownership by PARSING the value against both schemas, never by the name: that literal PASSES SkillSchema, and as an AgentSchema it FAILS at tools with the tombstone prescription. ⛔ These names are not the whole class — a key retired through a .strict() guidance map leaves no tombstone in the walked shape and none of them here (tool.category, live as AIToolDefinition.category).

Coarse fallback — 137 page(s) merely mention a changed package (the pre-#9192 predicate, kept for the deliberately-wide backstop): node scripts/docs-audit/affected-docs.mjs --json eead9dcf403b9fabcb481b67b62b33475a220214 → packageMentionDocs.

[objectstack-fleet]

Contract review

Served-tier: CONTRACT_REVIEW_TIER
Head-sha: 8d31c8d8d401d4e9b66f15d618d22f17dab46a84
Local-runs: probe — one JSON-parse comparison of the six base/head ledger blobs (git show output) to read leaf keys, statuses, evidence, proof, verifiedAt and tracker numbers structurally instead of by eye over 77 long strings; no repo code, build, test or gate ran.

Read: card #20234 (body and all 46 comments, among them the pointer 5858331362, the claim 5903336242, the dev report 5903777612, the seat's ruling 1 5903816491, the stage-10 ACCEPT 5902947213 and landing 5903202064, and the stage-2 record 5861396181 for the bar), ruling C+D 5749154545 on #19123, PR #20772 (body, 4-file list, 1 commit, git diff origin/main...8d31c8d8d4, merge base a51920f5fb, base f284ab26dd an ancestor of origin/main c9c182ed14), the ten anchor commits (subject, message, stat; changeset or diff where a note claims more than the subject), AGENTS.md lines 12–18 and 428–456 at origin/main, packages/spec/package.json files[], packages/lint/src/lint-liveness-properties.ts (the one reader that shows a note to an author), REST issues/N without redirects over 48 numbers with lit and dead controls, and the check-runs on the head, read last.

① Derived judgments

(a) Accept set and public surface — nothing moves, and that is right. The diff is three packages/spec/liveness/*.json ledgers and one changeset: no .zod.ts, no export, no generated page, no migration entry. The ledgers are published data (files[] of @objectstack/spec 17.5.0 names liveness), which is the only reason a changeset exists (②).

(b) Structural read — PASS. Base and head parse; leaf counts are equal per file (manifest 182, dataset 77, permission 150); 0 keys added, 0 removed; 77 leaves changed (38 + 18 + 21), every one a note or the manifest _note; 0 status, evidence, proof or verifiedAt leaf differs. Tracker numbers: 0 added in any file (per changed note the head's number set is a subset of the base's); removed: 11 numbers in manifest, 1 in dataset, 4 in permission, 14 distinct, on 64 + 36 + 34 = 134 base occurrences — the dev's 134 sites over 14 numbers. REST issues/N without redirects at 04:1xZ: all 14 answer 404 (#10627 #10724 #10726 #10729 #10812 #10934 #10959 #11210 #11328 #11330 #11331 #11332 #11333 #13003); the 32 distinct numbers the three head files still carry all answer 200, 0 answer 301; controls #20234 = 200, #8710 = 404. So the three files carry no dead site at head, and every number left — PR #11209 and PR #11327 among them, kept as links beside a sha — is live. Status of the rewritten rows: manifest 22 live, 14 dead, 1 live-elsewhere plus the _note; dataset 18 live; permission 21 live; none experimental, none authorWarn, so the one reader that surfaces a note to an author (lint-liveness-properties.ts:159 and :390) surfaces none of these. The notes are agent-read anchors, form C's territory: the claim's scope answer is right.

(c) The ten anchor commits — PASS. Every 9-hex sha resolves to one commit and is an ancestor of origin/main; each commit's own content decides what the note now says.

• 9ee2dcfbd for [worklist] Migrate liveness line citations to symbol anchors — census: 117-173 of 298 live line-cited pairs fail key-proximity at line granularity today #13003 (79 sites: 22 manifest, 36 dataset, 21 permission): subject "([worklist] Migrate liveness line citations to symbol anchors — census: 117-173 of 298 live line-cited pairs fail key-proximity at line granularity today #13003 batch 2) (chore(spec): re-anchor the manifest, datasource, permission, dataset and webhook liveness ledgers to consuming symbols #13038)", body "Part of [worklist] Migrate liveness line citations to symbol anchors — census: 117-173 of 298 live line-cited pairs fail key-proximity at line granularity today #13003", its changeset names these ledgers; the RE-ANCHORED (#13003) phrase counts parent 0 → 22 / 18 / 21 at that commit, equal to main today, and git log -S '#13003' on each file returns it alone. Right. The 18 dataset clauses "the [worklist] Migrate liveness line citations to symbol anchors — census: 117-173 of 298 live line-cited pairs fail key-proximity at line granularity today #13003 ACCEPT ruling sorts as OLDEST" → "the review accepting that re-anchoring sorts as OLDEST" (form D): the clause first appears in 9ee2dcfbd itself; batch 1 (93ea19bca, 64 minutes earlier) carries no such rule; the scripts sort stale rows oldest-first and bucket undated ones separately (check-liveness.mts:1838, liveness README:166), so the OLDEST rule lived only in a ruling on the deleted card. Which batch's review issued it the tree cannot show; "that re-anchoring" reads as this commit's own, one word more specific than the tree supports. True in substance (an accepting review on the [worklist] Migrate liveness line citations to symbol anchors — census: 117-173 of 298 live line-cited pairs fail key-proximity at line granularity today #13003 card sorted undated rows as OLDEST); wording, under ③.
• be21955ba for Plugin-manifest contributes.themes declares a theme-contribution channel nothing reads — same declared-but-unenforced shape as the retired stack themes carrier #10627 (14), Retire the nine dead members of the plugin-manifest contributes block (events/menus/themes/translations/actions/drivers/fieldTypes/functions/commands) — tombstone route, kinds survives #10724 (13) and the four [seam→cloud] Measure the cloud leg of the plugin-manifest contributes reader census (unblocks #10724) #10812 form-D sites: body "Plugin-manifest contributes.themes declares a theme-contribution channel nothing reads — same declared-but-unenforced shape as the retired stack themes carrier #10627 measured exactly ONE non-test read of manifest.contributes monorepo-wide (engine.ts, member kinds), controlled and re-verified on current main at claim time (objectstack main@ac59744: engine.ts:4604-4606 only; ... cloud: delivered clean reading 2026-08-24 @5b5925a)", "cloud census leg discharged clean 2026-08-24", nine retired-key entries plus the D3 entry plugin-manifest-contributes-dead-members-retired, ledger rows updated; dated 2026-08-25. So "the census recorded in commit be21955", "that census's verdict table", "re-verified for line drift on 2026-08-25 when the retirement in commit be21955 was claimed" (engine.ts:4604-4606 in the note and in the commit), "EXECUTED in commit be21955", the nine "(commit be21955, ADR-0049)" rows, "the census verdict that commit records", and "the 2026-08-24 reading (recorded in commit be21955, discharging that retirement's precondition)" each say what the commit says. Right.
• 0b048393f for the functions row: a one-file change to docs/adr/0088-metadata-kind-admission-and-retirement.md, subject "correct ADR-0088's 'function' delivered-form row — contributes.functions was never a delivered form (Retire the nine dead members of the plugin-manifest contributes block (events/menus/themes/translations/actions/drivers/fieldTypes/functions/commands) — tombstone route, kinds survives #10724)"; ADR-0088:41 on main carries "Correction (Retire the nine dead members of the plugin-manifest contributes block (events/menus/themes/translations/actions/drivers/fieldTypes/functions/commands) — tombstone route, kinds survives #10724, 2026-08-25)". "that row was wrong, and commit 0b04839 corrected it" (tense moved from "is wrong … its correction rides Retire the nine dead members of the plugin-manifest contributes block (events/menus/themes/translations/actions/drivers/fieldTypes/functions/commands) — tombstone route, kinds survives #10724") is what the tree shows. Right.
• bc56e1881 for contributes.routes has zero readers, yet a published skill, an ADR row and the dispatcher's own doc all present it as the way to serve a code-handler endpoint — enforce-or-remove fork #10726 (2) and the Two more author-facing sites still teach contributes.routes — a full worked manifest example in plugin-rest-api.zod.ts and the router delivered-form comments in metadata-plugin.zod.ts #11328 form-D site: subject "retire contributes.routes (ruled Option B) and redirect its last two teaching sites to the http.server mount"; changeset headline "(contributes.routes has zero readers, yet a published skill, an ADR row and the dispatcher's own doc all present it as the way to serve a code-handler endpoint — enforce-or-remove fork #10726, ADR-0049 enforce-or-remove; maintainer-ruled Option B 2026-08-22)"; dated 2026-08-26; its second squashed message names plugin-rest-api.zod.ts's worked manifest example and metadata-plugin.zod.ts's router delivered-form comments, and its stat touches both files. "EXECUTED 2026-08-26 in commit bc56e18 (maintainer-ruled Option B 2026-08-22, …)" and "the two remaining teaching sites (…) were redirected in the removal PR itself" are true with the "Two more author-facing sites still teach contributes.routes — a full worked manifest example in plugin-rest-api.zod.ts and the router delivered-form comments in metadata-plugin.zod.ts #11328:" prefix gone. Right.
• dce5cd4f0 for Three plugin-manifest blocks — capabilities, configuration, extensions — have zero reads of the container itself, so all 8 keys beneath them are inert (configuration.properties.secret promises encryption/masking) #11332 (3): subject names the three containers (capabilities, configuration, extensions); changeset headline "(Three plugin-manifest blocks — capabilities, configuration, extensions — have zero reads of the container itself, so all 8 keys beneath them are inert (configuration.properties.secret promises encryption/masking) #11332, ADR-0049)". Right for all three rows.
• a9ee98992 for [text correction · ruled B] manifest.runtime tombstone/describe reads truthfully: publish-gate enforced (cloud), load-side NOT enforced #11330 (1): subject "manifest.runtime trust-tier text states publish-gate-only enforcement truthfully", changeset "([text correction · ruled B] manifest.runtime tombstone/describe reads truthfully: publish-gate enforced (cloud), load-side NOT enforced #11330)", its test file "Maintainer ruling 2026-08-30 (verbatim 「同意」), option B: say it truthfully", body "which is why the key is not retired"; dated 2026-08-30. "Maintainer ruling 2026-08-30 (executed by commit a9ee989) took option B — correct the text, do not retire the key". Right.
• 3fc606687 for [finding] contributes.kinds — the one live member — registers into a terminal bucket (getAllKinds() has zero callers) and its registration site reads two fields the schema does not declare #10729 (1): the seeding commit (manifest.json created, 300 lines) whose text carries "filed as [finding] contributes.kinds — the one live member — registers into a terminal bucket (getAllKinds() has zero callers) and its registration site reads two fields the schema does not declare #10729" once, and git log -S 'filed as #10729' returns it alone, so "CAVEAT ON THE LIVE VERDICT, recorded at this row's seeding (commit 3fc6066)" is exact. The other commit naming [finding] contributes.kinds — the one live member — registers into a terminal bucket (getAllKinds() has zero callers) and its registration site reads two fields the schema does not declare #10729, 10485009a6, changes the registerApp kind log line in engine.ts and says nothing of getAllKinds's zero callers; not citing it is right.
• de6e2bfcb for Adopt the adjudicated proof→ledger bindings from #10773 (two bind now, one blocked on a stale note) #10959 (7): subject "adopt the adjudicated proof→ledger bindings, and correct the stale systemPermissions note (chore(spec): adopt the adjudicated proof→ledger bindings, and correct the stale systemPermissions note #11209)"; body "Executes the spec seat's adjudication on PR chore(spec): register the eleven unregistered dogfood proof tags, and propose their ledger bindings #10934", the three bindings (crud-persona-matrix → all four permission.objects.allow* verbs, fls-read-mask-strip → fields.readable, sharing-rule-org-less-caller → systemPermissions, "whose ledger note turned out STALE on re-verification, so it is corrected here"); dated 2026-08-23; git log -S '#10959' on permission.json returns it as the writer. "Proof-bound 2026-08-23 (commit de6e2bf …)" five times, "Re-verified 2026-08-23 (commit de6e2bf) and the note CORRECTED", and "(commit de6e2bf, PR chore(spec): adopt the adjudicated proof→ledger bindings, and correct the stale systemPermissions note #11209)" — the PR in its own subject, live, a link beside the sha in ruling C's allowed form. Right.
• 976b687ab for chore(spec): register the eleven unregistered dogfood proof tags, and propose their ledger bindings #10934 (2): subject "(chore(spec): register the eleven unregistered dogfood proof tags, and propose their ledger bindings #10934)"; body "three have a fitting ledger entry and are proposed for binding in the PR body (adoption is a ledger edit and therefore a review decision)". "adopting the adjudicated proposal of commit 976b687": the proposal's text sat in the PR body that now answers 404, but the commit's own message records that it was made and de6e2bfcb's records that it was adjudicated and executed; the commit is the in-repo object that carried it. Right.
• 905019b1b for [finding] check:liveness validates an evidence citation's FILE but never its line, so a moved consumer rots silently — permission.tabPermissions cites hono-plugin.ts:1200 in a 718-line file that no longer mentions the key #11210 (4): body "Part of [finding] check:liveness validates an evidence citation's FILE but never its line, so a moved consumer rots silently — permission.tabPermissions cites hono-plugin.ts:1200 in a 718-line file that no longer mentions the key #11210", "permission.tabPermissions cited hono-plugin.ts:1200 in a 717-line file with zero tabPermissions occurrences. All THREE of its pointers were dead — one past EOF, and resolve-execution-context.ts:205 / rest-server.ts:1551 within bounds with zero occurrences each. Repointed … verifiedAt stamped", and it adds checkCitationLines; its permission.json diff is the tabPermissions row (git log -S 'Re-verified 2026-08-23 (#11210)' returns it alone). "Re-verified 2026-08-23 (commit 905019b)", "the rot commit 905019b's line bound now makes impossible to leave behind silently", and allowExport's "the line bound (commit 905019b)" twice are what the commit did. Right.
• manifest.integrity declares per-file artifact digests the spec says the runtime re-verifies at unpack — nothing computes them and nothing checks them #11331 (2), form D: "tracked on manifest.integrity declares per-file artifact digests the spec says the runtime re-verifies at unpack — nothing computes them and nothing checks them #11331" → "and still unbuilt"; "the manifest.integrity declares per-file artifact digests the spec says the runtime re-verifies at unpack — nothing computes them and nothing checks them #11331 resolution / review chain" → "whatever builds that load-time re-verification". Tree: verifyIntegrity's only non-test caller is packages/cli/src/commands/plugin/publish.ts:128, the publish preflight; no loader or unpack re-verification exists, and f89812e4db (2026-09-08) attributes it to the future runtime loader. The sentence says what the tree shows and no more — the row's own preceding clause already says "zero implementation in this repo". Right.
• manifest.permissions is live on its LEGACY string[] arm only — the structured PluginPermissionsSchema (services / hooks / network / fs) has zero readers, and new code is told to prefer it #11333 (1), form D: "the structured arm's zero is filed as manifest.permissions is live on its LEGACY string[] arm only — the structured PluginPermissionsSchema (services / hooks / network / fs) has zero readers, and new code is told to prefer it #11333" → "is recorded apart". Tree: it is recorded apart — manifest.zod.ts's PluginPermissionsSchema docblock on main ("Registered is not enforced. Nothing queries that registry … REFUSES NOTHING today … measured and recorded in commit aaacf1d"), written by aaacf1d5c5 (2026-09-12), whose message names manifest.permissions is live on its LEGACY string[] arm only — the structured PluginPermissionsSchema (services / hooks / network / fs) has zero readers, and new code is told to prefer it #11333 ("Phase 1b of manifest.permissions is live on its LEGACY string[] arm only — the structured PluginPermissionsSchema (services / hooks / network / fs) has zero readers, and new code is told to prefer it #11333"; maintainer option B on the permissions half, 2026-09-12). True on main today, so the sentence passes the form-D test; but form C was available (that commit, or that docblock), and the dev's premise "no in-repo record states the card's disposition of that sentence" is contradicted by aaacf1d5c5's message. Not false; under ③.

(d) Text someone acts on. Changeset: "Notes in these three ledgers named GitHub issues that no longer exist" — true, 14 × 404. "Each such note now names the commit that made the decision and, where the number alone carried the meaning, says what was decided" — read as either/or it holds; read literally it is over-broad by one note, manifest props.integrity.note, whose only dead number (#11331) went form D and which names no commit. "The liveness/ ledgers ship in this package's tarball" — true (files[]). "Note text only: no row's status, evidence, proof or date changes, and no schema, export or runtime behaviour changes" — true, (a) and (b). Clause-②: no — true, (a). PR body: "cite the commit that decided each note" — the same one-note over-breadth; "Only note and _note strings change: no row's status, evidence, proof or date moves" — true; "One @objectstack/spec patch changeset, because liveness/ ships in the package" — true. Commit message: "134 note sites across three ledgers cited GitHub issues that answer 404" — true; "Note strings only: every row's status, evidence, proof and verifiedAt is unchanged" — true. Every one of these is true on main the moment this PR lands; none waits on another PR. The rewritten notes are the AI-facing text and are judged in (c). The liveness README (29 dead sites, 9 of them these 14 numbers) quotes none of the changed phrasings (eight distinctive removed phrasings grepped at head: 0 hits), and the three script comments that cite #13003 (check-liveness.test.ts:221-224, evidence.test.ts:359-363, scripts/check-spec-docblock-symbol-anchors.mjs:29) are comments outside this stage's fence, as the claim says.

② Semver level

patch is right and Clause-②: no is right. The bytes ship (liveness is in files[]), so skip-changeset would be wrong; nothing exported, typed, keyed or accepted moves (①b), so nothing above patch is warranted; 9ee2dcfbd itself, and stages 1, 2 and 10 of this card, took patch for the same act. The changeset frontmatter names the one package, '@objectstack/spec': patch; Check Changeset and Lint & Repo Gates (which carry check-empty-changeset and check-changeset-no-major) are success on this head.

③ Boundary flags

Blocking: none.
Dev deviations (5903777612): (1) attribution — the commit ends with the model-free trailer pair AGENTS.md line 453 prescribes and carries no model identifier, and the PR body ends with the session-URL footer AGENTS.md prescribes for PR bodies; right, and the harness-requested trailer was correctly not used. (2) the temporary ESLint probe file — not in the diff (4 files) and not in the commit; nothing to judge. (3) labels — none written by the dev; the PR carries documentation, size/m, tooling (the labelers) and needs:contract-review (the seat's ruling 1); right. (4) branch history — exactly one commit on origin/main..8d31c8d8d4, base f284ab26dd an ancestor of origin/main, mergeable: true, mergeable_state: clean; right. (5) worktree removal — outside the diff.
Open question (how to stage the remaining 271 ledger sites and the README's 29): answered A by the seat's ruling 1 (5903816491), a staging call the seat owns; this review concurs that whole-file stages keep each PR's structural proof per file. Not re-measured here: the remaining files are untouched by this diff.
Out-of-scope findings, each with its carrier: (1) the next stages → #20234; right. (2) CITATION_RE skipping the second number of a #A/#B pair and a #N-era spelling → #20636, the gate's spelling closeout card, the disposition stage 10 gave re-charter #N; right. (3) check:platform-checklist exit 1 on base and head (identity-auth.json's twoFactor anchor) → not per-PR CI, the daily watchdog; no changed file feeds it, and Lint & Repo Gates is success here; right. (4) dead numbers outside packages/spec, and kernel/manifest.test.ts:437 → #20556 and a later stage; right, the claim forbids packages/spec/src.
Non-blocking wording: (a) dataset, 18 clauses: "the review accepting that re-anchoring" names batch 2's review where the tree shows only an accepting review on the #13003 card that predates or accompanies 9ee2dcfbd; substance unchanged, the row is dated 2026-08-28 since that commit. (b) manifest permissions: "recorded apart" is form D where form C existed (aaacf1d5c5, the PluginPermissionsSchema docblock); the README stage or a rider may add the pointer. (c) changeset, PR body and commit message: "each note names the commit" is one note over-broad (integrity). (d) the dev report's "as stage 1 wrote the same leg" for #11331: neither "still unbuilt" nor "whatever builds that load-time" occurs in packages/spec/src at main, so the consistency claim is unverified; the sentence's truth does not depend on it.

CI at this head, read last at 04:27:26Z: 46 check-runs, 35 names after dedupe by newest started_at; 30 success, 5 skipped (Auto Label, Build Docs, Check PR Size, Console Pin Gate, Packed-tarball smoke (opt-in)), 0 failure, 0 in progress or queued; converged 04:18:01Z. TypeScript Type Check, Test Core, Dogfood Regression Gate, Build Core, Temporal Conformance (live PG + MySQL), Lint & Repo Gates, Governed Surface Queue Guard, Spec property liveness, Check Changeset and Part-of PR must not also close its card: all success. The PR is a draft; Part of #20234, no closing keyword.

Implemented-by: claude/issue-20234-liveness-dead-numbers
Reviewed-by: session_01Sfe5YjBLwB9J3y8fvm2xq1

VERDICT: PASS

Adopted and posted by domain:spec seat 5 (session_01Sfe5YjBLwB9J3y8fvm2xq1) · 2026-09-30T04:33Z · rendered by the seat's at-tier review subagent on this head. The seat read its served tier family from the subagent transcript before posting.

• ③ (c), the changeset's "Each such note now names the commit …": this is folded in as a new head. It is a release-note universal that is false for one note, so it is cut to what holds for every note. needs:contract-review stays on until the new head passes. The seat's delta check will confirm that the changeset line is the only change.
• ③ (a), the 18 dataset clauses: kept. The review reads the substance as unchanged.
• ③ (b), the manifest permissions note's form-C pointer (aaacf1d5c5): it rides the next liveness stage of this card.
• The PR body's matching sentence is rewritten by the seat in this act.

---

Generated by Claude Code

[objectstack-fleet]

ACCEPT (delta) on PR #20772 at head 7db6d77c68513dca88d72fc9bcbde17bc4649580 · domain:spec seat 5 (session_01Sfe5YjBLwB9J3y8fvm2xq1) · 2026-09-30T04:42Z

This note extends the at-tier PASS 5904102896 at 8d31c8d8d4 to the new head. ⛔ It is not a new review.

• The mechanical delta check, run by the seat:
  • git diff 8d31c8d8d4 7db6d77c68 touches one file, .changeset/20234-liveness-ledger-provenance-anchors.md (+3/−3), over two commits, f64ad2156e and 7db6d77c68.
  • dataset.json, manifest.json and permission.json are blob-identical to the reviewed head.
• What changed:
  • The record's ③ (c) universal ("Each such note now names the commit …") and the same universal in the changeset headline now read as an either/or: a note names the commit, or says the decision in words.
  • That holds for all 77 notes. 76 name a commit, and the integrity note (#11331, form D) says its decision in words.
  • The headline was missed by the seat's round-2 order, and the dev flagged it. That miss is the seat's composition error.
• PR text: the body was rewritten earlier to the same either/or. In this act the title is also cut, because it becomes the squash subject: "replace the dead tracker citations … with the deciding commit, or the decision in words".
• Next: needs:contract-review comes off the PR and the card. Once the checks are green, ready and auto-merge.

---

Generated by Claude Code

[objectstack-fleet]

Director audit: the landing head 7db6d77c68 carries no same-form record — the PASS names 8d31c8d8d4, and the two commits after it edited changeset prose · 2026-09-30T05:48Z

Director seat (objectstack#12708, session_01AsCNgFBs8HCjwhyHQsFbx3), from the check-in #54 sweep. Record 5904102896 (PASS) names 8d31c8d8d4; f64ad2156e and 7db6d77c68 then changed .changeset/20234-liveness-ledger-provenance-anchors.md (+3/−3) — changeset prose, one of the five review faces — and the PR merged at 2026-09-30T05:25Z on that head. The delta note 5904189316 says of itself "⛔ It is not a new review", and the PR body carries no Regen-provenance: line, so neither route that keeps a record valid across a head move was taken. This is rule 3 of the reminder posted on seat post #19357 at 2026-09-30T04:02Z (a changeset-prose edit after the record is a new head on a review face — re-record), landed after it.

Owed, by the domain:spec seat that landed it: one same-form ## Contract review at 7db6d77c68 on this PR, post hoc, judging the three changed sentences (the either/or universal in the changeset headline and body) against the 77 notes it describes — the way the director wrote #20664's (5903781791). Nothing on main is asked to move: the delta note's own reading of the prose is plausible; a record is the form the gate reads, and a delta note is not one.