test(spec): the six reader-bound needles move with their readers, naming each cited decision in words instead of a tracker number (stage 31)

[objectstack-fleet]

Part of #20749
Clause-②: no

Stage 31 of this card, the needles stage: answer A in the seat's ACCEPT 6074213330, under the claim 6084058576. Six test strings carried a tracker id that a reader matches. Each one moves with its reader in this PR, and the match stays exactly as strong: the new words sit in the same reader places the old id did. Three class (e) strings that regrew since stage 30 drop a number: two titles that already explain it, and one ledger why: value that keeps its ruling record id (the claim amendment in ACCEPT 6085085283). Text only: no schema shape, key, .describe() string or behaviour moves.

Census, re-taken at the base

Stage 28's census28.cjs, byte-identical (md5 31d8488b…), calibrated first against nine published readings, all reproduced exactly: 128/130, 191/200 and 73/76 (stages 27 to 28), 73/76 and 16/17 (stage 29's base and head), 17/18 and 10/10 (stage 30's base and head), 17/18 at e75dceddd and 10/10 at stage 30's landing 961d365f1.

At the base e148ca984: 13 messages / 13 ids in 10 files (titles 2/2, other 11/11).

• the 6 reader-bound strings (moved here);
• the 4 CSS hex colours that stage 21's ACCEPT kept (ui/dashboard-chart-structure-refusal.test.ts:94 x2, ui/dashboard.test.ts:124 x2);
• 3 regrown since stage 30's landing:
  • 2 titles from PR 22315 (automation/end-node-outcome.test.ts:67, automation/io-node-config.test.ts:472), rewritten here;
  • 1 ledger why: value from PR 22425 (system/metadata-form-zod-reconciliation.test.ts:497). The claim amendment in ACCEPT 6085085283 admitted it, and the patch round rewrote it: #22211 A becomes letter A, and the ruling record id it cites stays.

At the head a766d439f: 4 / 4 in 2 files, the 4 hex colours only, which is the card's end state. Titles 0/0. (fbec38c32, before the patch round, read 5 / 5 in 3 files.)

No file is deferred. At 15:45Z there were 11 open PRs (418 file rows, each list complete against the PR's changed_files). None touches any of the 14 files in this diff, or the 3 files that still carry an id. The lit control, ui/component.zod.ts, is found in 2 of them (22421 and 22496).

The six pairs

Each needle and every reader place it matched move together, one span per line. "Places" are the reader lines where the needle matched inside the exact text the test reads.

pair	test string	reader	needle, base → head	reader places, base → head
P1	ai/build-progress.test.ts:236	ai/build-progress.zod.ts, the whole file (SOURCE)	cloud#2172 → the cloud build-progress card	8, 27, 77, 84, 108, 123, 177 → the same seven
P2	ai/build-progress.test.ts:237	the same file	objectui#7388 block 2 → block 2 of the objectui build-panel card	85 → 85
P3	contracts/approval-service.test.ts:274	the continueRestoredRun docblock in contracts/approval-service.ts	#16495 → the automation-service contract	999 → 999
P4	ui/notification.test.ts:123, plus its own anti-vacuity regex at :134	the tombstone note label in ui/notification.zod.ts	// [#4610] → // [dual-source removal] (and the regex ^\[#4610\] → ^\[dual-source removal\])	94 → 94
P5	ui/strictness-batch14.test.ts:395	ui/notification.zod.ts and ui/sharing.zod.ts, each whole	#5015 → no-door retirement	48, 103 and 24, 114 → the same
P6	ui/component-props-unknown-members.pin.test.ts:322 (the opaque line's ruling)	the assertion at :417	#21704 → the five members the last typing stage stopped on (the regex follows)	the one ruling value → the same

What each number meant, now said in words:

• P1, P2. The cloud card is the one that ruled the build-progress phase vocabulary in and owns where the emitter sits; the docblock says both. Block 2 of the objectui card is the piece that admits verify in extractBuildProgress, the consumer's strict parse.
• P3. The sibling this verb copies is IAutomationService.cancelRun / restoreConsumedSuspension, as the automation-service contract declared them.
• P4. The ./ui Notification / NotificationConfig wrappers were removed as dual-source, and the note's label now says so.
• P5. NotificationActionSchema and EmbedConfigSchema were retired because they had no authoring door (ADR-0049 enforce-or-remove). Both modules record that retirement in the same four places as before.
• P6. The decision card is the one on the five ComponentPropsMap members the last typing stage stopped on. The record id 5979239990 stays.

Readers beyond the claim's "about" lines. The claim lists build-progress.zod.ts about :8, :27, :85 and sharing.zod.ts about :24. The old needles also matched build-progress.zod.ts 77, 84, 108, 123 and 177, and sharing.zod.ts 114. Leaving any of those would have made the new needle match fewer places than the old one, which is a stronger test, not the same one. So they move too, inside the two files the claim admits. Nothing outside the claim's files moves except the two generated pages, both written by gen:docs only: ai/build-progress.mdx, as the claim names, and ui/sharing.mdx, which renders the sharing.zod.ts module docblock.

Shipped text that moved. This is listed for the review.

• build-progress.zod.ts: the module docblock (lines 8, 27, 77, 84, 85), the BUILD_PROGRESS_PHASES docblock (108, 123) and the hop field docblock (177).
• approval-service.ts: the continueRestoredRun docblock (999), which dist/contracts/index.d.ts carries.
• notification.zod.ts: comment lines 48, 94 and 103.
• sharing.zod.ts: module docblock line 24 and comment line 114.
• The ai/build-progress reference page's description: front matter now ends with "(the cloud build-progress card, ruling A)". page-description.ts strips only a parenthetical that holds a tracker or ADR id, so the words now stay in the description.

No .describe() string moves.

Ids that stay on those lines on purpose. These are comment citations that no test matches, so they are not needles: #15389 and decision batch #106 beside P3, and the other ids in the same docblocks.

Class (e) regrowth

(#22110) is dropped from the two titles. Each already states that card's decision: a flow text slot reads {{ }} holes and refuses a single-brace token. Read with the card's 17 comments and its landing record. There is no reader: a git grep -F of each old title and of each distinctive head finds only the literal itself. The lit control composeStacks is found in stack.zod.ts; the dark control exits 1.

Text only, proven

• Byte level, all 13 edited files. subst31.cjs applies the plan to the base text and compares the result with the head. The plan was written at 15:48:21Z, before any edit: 31 spans, each required to occur exactly once on its base line. The result is 13/13 SAME, line counts unchanged, including both generated pages and the description: line exactly as predicted. Every old span carries an id, and no new span carries # plus digits.
• Controls, 9/9 as predicted. One character in an unplanned comment, an id appended inside a new span, a reverted span, an inserted line, trailing whitespace, a generated page edited off-plan, a plan with an absent span and a plan with an id in a new span all go red. The unmodified copy reads SAME.
• Stage 28's textonly28.cjs, byte-identical, on the 7 test files: 5 SAME (2 title, 5 declared changes). notification.test.ts and the component-props pin stop at their regex reader lines (134, 417), because that instrument masks string tokens, not regex literals. With only those two lines put back to base, both read SAME (1 declared change each).
• Per-file test counts and status sequences, base vs head, 7 files, --project local --project repo, JSON reporter: 193 / 193, identical in 7/7. Two fullNames change (the two titles) and 0 differ from the plan. There are 0 duplicate fullNames on either side. Names carrying # plus digits: 2 at base, 0 at head.

Same strength, proven

Places. sites31.cjs finds every place each needle matches, inside the exact text its test reads. The old needles' places at base equal the new needles' places at head, pair by pair (the table above). Each new needle has 0 places at base, and each old needle has 0 at head.

Mutations. Each leg goes through scripts/ablation-replace.mjs, which requires the anchor count, restores to HEAD and proves blob == HEAD after every leg. The test reads the source with readFileSync from src, so no build sits between the mutation and the test. The predictions were written at 15:55:52Z, before the first run. The same matrix ran at the head and, with the worktree detached at the base, at the base:

drift	base	head
P1, every place	red	red
P1, one place (84) of seven	green	green
P2	red	red
P3	red	red
P4, the label (the anti-vacuity test and 4 more go red)	red	red
P5, both places in notification.zod.ts	red	red
P5, both places in sharing.zod.ts	red	red
P5, one place (48) of two	green	green
P6, the assertion text	red	red
P6, the ruling value	red	red

• Result: 10/10 at the head and 10/10 at the base, as predicted. Every red fails on the needle assertion itself.
• The null legs: the first P2 leg at the head never ran. ablation-replace refused it, because its replacement was a substring of the anchor. The base P2 leg as first written had the same defect and was not run. Both were re-anchored with a replacement that is not a substring, and that amendment was written down before the re-run, with the prediction unchanged.
• Restore: after the matrix the worktree is back on the branch at fbec38c32, and git status is clean.

Verification at fbec38c32

All heavy runs went through scripts/pm/os-verify-lock.sh, slot dev-20749-s31. Each exit was written to a file or read from the lock's VERDICT line.

• Workspace build: pnpm turbo run build --concurrency=2 over ./packages/* and ./packages/*/* gives 71/71 tasks, exit 0. It is the prerequisite for the gates that read dist.
• Spec typecheck: exit 0, and check:test-typecheck holds 52 files / 246 errors / 135 pinned signatures, unchanged. tsc --listFilesOnly -p tsconfig.test.json holds all 7 edited test files (617 test files in the program).
• Spec unit tier, --project local, in full: 630 files passed, 18798 tests passed and 1 todo, exit 0.
• Spec repo tier, --project repo, in full: 54 files passed, 915 tests passed, exit 0.
• check:generated: all 15 generated artifacts are up to date. Before gen:docs ran, it named only content/docs/references/** as stale.
• Derived gates: dispatch-gates --commands derives 109 commands from the merge base e148ca984 (14 paths). All 109 exit 0, and --ran reconciles 109 derived, 109 run, 0 NOT-MEASURED, a derived zero.
• Roster families: the 6 whose roster sits in a directory this diff touches also exit 0 (check-changeset-fixed, check:meta-url-spelling, check:spec-changes, check:authz-resolver, check:error-code-casing, check:filter-alias-parity).
• ESLint, a proven narrowing: --no-inline-config over the 11 changed .ts files gives 0 errors and 0 warnings. ESLint's own config reports 11 configured, 0 ignored and 0 with type-aware parser options, so the diff cannot move a verdict on an untouched file. The 3 Markdown files are outside ESLint's config.
• Control bytes: the scan over the 14 changed files finds 0. check:nul-bytes (in the derived set) exits 0.

Patch round at a766d439f (open question answered A, ACCEPT 6085085283)

One span on one line, system/metadata-form-zod-reconciliation.test.ts:497: (ruling record 6070963704, #22211 A): becomes (ruling record 6070963704, letter A):. The commit is a766d439f, on top of fbec38c32, with no merge of main (the PR reported no conflict, and git merge-tree onto origin/main 446c8b2a6 is clean).

• Text only, byte level: the one-span plan, written at 16:36:23Z before the edit, applied to fbec38c32, equals the head: SAME, 1616 lines on both sides. A control that re-adds an id inside the span reads DIFFERENT. Stage 28's textonly28.cjs with line 497 declared reads SAME (552 string tokens, 1 changed, declared); without the declaration it refuses the line, as it should.
• No reader: a git grep -F of the old text, and of the value's opening words, finds only the literal itself. No assertion reads the text, only its length (more than 20).
• That file's tests: 76 / 76 passed at base and head, under --project local --project repo. The status sequence is identical and no fullName changed.
• Census: census28.cjs at the head reads 4 / 4 in 2 files, the 4 named hex colours. The same instrument reads 5 / 5 at fbec38c32 and 13 / 13 at the base e148ca984.
• Full runs at a766d439f:
  • workspace build: 71/71;
  • spec typecheck: exit 0, with check:test-typecheck at 52 / 246 / 135 held, and the patched file is in the tsconfig.test.json program;
  • spec --project local: 630 files, 18798 passed, 1 todo;
  • spec --project repo: 54 files, 915 passed;
  • check:generated: all 15 up to date.
• Gates: dispatch-gates --commands derives the same 109 commands, all exit 0, and --ran reads 109 / 109 / 0 NOT-MEASURED. The 6 directory-sharing roster families also exit 0. ESLint over the file: 0 / 0; it is configured, not ignored and not type-aware.
• CI on a766d439f at 17:05Z: 33 success, 2 skipped, none red. On fbec38c32, the push of a766d439f cancelled 14 jobs, and the aggregate TypeScript Type Check job failed at its "Verify every type-check lane succeeded" step because its lanes were cancelled. That red is on the superseded head only.

Changeset

@objectstack/spec patch: shipped text moved. The package's files ships src/**/*.zod.ts, and the continueRestoredRun docblock is in dist/contracts/index.d.ts. npm pack --dry-run: 2069 files, 0 *.test.ts. The three edited .zod.ts sources ship, and dist/contracts/index.d.ts carries the new docblock words (the old ones are in 0 dist files). Clause-②: no: no accept-set change.

Acceptance notes

• Kept, by ruling: the 4 CSS hex colours (stage 21's ACCEPT).
• The regrown ledger why: value (system/metadata-form-zod-reconciliation.test.ts:497, added by PR 22425 after stage 30) was admitted by the claim amendment and rewritten in the patch round. The census now reads only the 4 hex colours.
• Comment citations in the four reader files that no test matches stay as they are (for example cloud#1838, objectui#7388 in other sentences, #4535, #5781, #15389).
• An observation, not filed. The build-progress liveness watch names the objectui card as one of two carriers. That card was closed on 2026-09-27 with its block 2 handed off and "not tracked here". The objectui changeset this repo now carries (.changeset/console-f0268ad78485.md) says the chat panel reads cloud's verification loop. The docblock's carrier sentence may be stale. This stage keeps what it says and only replaces the id.
• A wide reading, for the record. Counting any # plus digits, 4 titles carry PD #12 / batch #58 (conversions/conversions.test.ts, system/i18n-resolver.test.ts). These are two-digit labels below the census id pattern, the same at base and head, and outside the census population.

---

Generated by Claude Code

[github-actions]

📓 Docs Drift Check

This PR changes 1 package(s): @objectstack/spec, touching 2 documentable anchor(s). ⚠️ 2 changed file(s) yielded no anchor (packages/spec/src/ui/notification.zod.ts, packages/spec/src/ui/sharing.zod.ts), so the pages documenting them are NOT COVERED by this run — this is not a clean bill of health for those files.

1 hand-written doc(s) NAME something this change touched and may need an implementation-accuracy re-verification:

• content/docs/ai/actions-as-tools.mdx (via IApprovalService (symbol, a top-level interface))

What this run could not see

• 2 changed file(s) yielded no anchor (packages/spec/src/ui/notification.zod.ts, packages/spec/src/ui/sharing.zod.ts) — pages documenting those are invisible to this run
• the SDK route bridge reached 54 of 206 client-bound route-ledger rows — the other 152 have no registrar path: tail to select them, so pages documenting THEIR client methods cannot appear above, on this or any run. Of those 152: 0 are remediable by widening that discovery convention (an in-repo file declares the path; the convention did not scan it); 55 are structural — on a ledger where NOT ONE row is declared in-repo, so no discovery change reaches them at any price; 97 are undecided (no in-repo declaration, on a ledger that has other in-repo registrars — absence and an unreadable spelling are not distinguishable here). The rows themselves: node scripts/docs-audit/affected-docs.mjs --bridge-coverage
• 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 — 139 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 446c8b2a6420a61a2862e6f5140dda71a53316d1 → packageMentionDocs.

Which tree this was computed on

This run read content/docs from 81b61a597b8360d84aee075a2397a61e70e78b04 — the merge of head a766d439f48953f3404f6d3d7afed0cd2617ada1 into base 446c8b2a6420a61a2862e6f5140dda71a53316d1, which is what actions/checkout gives a pull_request run. Not the PR head.

A worktree cut from an older main holds a different content/docs, so re-deriving there can legitimately return a different list — that is a different tree, not a wrong row. To answer on the same tree:

# while this PR is open — GitHub drops the merge commit once it closes
git fetch origin 81b61a597b8360d84aee075a2397a61e70e78b04 && git checkout 81b61a597b8360d84aee075a2397a61e70e78b04
# afterwards, rebuild it from the two parents, which stay fetchable
git fetch origin 446c8b2a6420a61a2862e6f5140dda71a53316d1 a766d439f48953f3404f6d3d7afed0cd2617ada1 && git checkout -B drift-repro 446c8b2a6420a61a2862e6f5140dda71a53316d1 && git merge --no-ff a766d439f48953f3404f6d3d7afed0cd2617ada1

node scripts/docs-audit/affected-docs.mjs --json 446c8b2a6420a61a2862e6f5140dda71a53316d1

⚠️ That checkout carried uncommitted changes, so the commit above does not fully identify what was read.

Advisory only, and a precision-first one (#9192): a page is listed because it names a
symbol, wire route or SDK method this diff touched — not because it mentions a changed
package. Each row says which anchor put it there, so a wrong row is reportable rather than
merely annoying. To re-verify, run the docs-accuracy-audit workflow scoped to these files:
node scripts/docs-audit/affected-docs.mjs 446c8b2a6420a61a2862e6f5140dda71a53316d1 → pass the list as
args.docs, on the commit named under Which tree this was computed on.

[objectstack-fleet]

Contract review

Served-tier: CONTRACT_REVIEW_TIER
Head-sha: a766d439f48953f3404f6d3d7afed0cd2617ada1
Local-runs: none

PR #22511 (stage 31 of #20749, the needles stage): a draft against main, head repo equal to base repo, branch claude/issue-20749-stage-31-needles, 15 files, +48 / −32, merge base e148ca984. Inputs: the card body and all 130 comments (the stage-31 claim 6084058576, the ACCEPT 6085085283 that amends it, the two newest dev reports 6085042055 and 6085588181, stage 30's ACCEPT 6074213330 and landing 6074840110, and stage 20's 5998488373 and stage 21's 6001279159, which set the needles stage and kept the four hex colours), the PR body and file list, the net diff against main at the merge base, and the head's check-runs, read at 2026-10-09T17:19Z. Nothing was built, run, re-run or ablated here: every reading below is git show / git grep on the fetched head and base refs, or REST.

Check-runs on the head at this reading: 42 runs; 38 success, 4 skipped, 0 failed, 0 queued or in progress. The seven required contexts are green: Lint & Repo Gates, TypeScript Type Check (with its four lanes workspace, source gates, debt ledger, consumer gates), Test Core (shards 1/6 to 6/6 and the rollup), Dogfood Regression Gate (1/3 to 3/3 and the rollup), Build Core, Temporal Conformance (live PG + MySQL) and Governed Surface Queue Guard. Also green: Check Changeset (both runs), Spec property liveness, Build Docs, Check Documentation Links, Dogfood Verify CLI, Flag docs affected by code changes, Part-of PR must not also close its card (both), The card this PR closes must claim this branch (both), both single-writer guards, both same-issue guards, Auto Label, Check PR Size and filter. Skipped by design: Console Pin Gate and Packed-tarball smoke (opt-in); Auto Label and Check PR Size each also carry one skipped re-trigger beside their green run. No run is red. The cancelled-lane red the patch report names sits on the superseded head fbec38c32, not on this one.

① Derived judgments

• Accept-set: no change — right. Read on the diff: 0 schema lines, 0 .describe() strings, 0 keys, 0 defaults, 0 exports and 0 types move. Every changed non-test line is inside a /** */ or // comment: ai/build-progress.zod.ts (8 lines), contracts/approval-service.ts (1), ui/notification.zod.ts (3), ui/sharing.zod.ts (2); the rest of the diff is seven test files, two generated pages and one changeset. Type Check · source gates and Spec property liveness are green on the head.

• Public surface: shipped text moves and nothing else — right. packages/spec/package.json at the head ships src/**/*.zod.ts, so the three .zod.ts comment edits ship as source; approval-service.ts is not a .zod.ts, but its interface docblock is emitted into dist/contracts/*.d.ts. The two reference pages regenerate from the module docblocks, and check:docs (inside the green TypeScript Type Check) holds them equal to what gen:docs produces. That is why a changeset is owed (②).

• Meaning: every shipped text keeps what it says and only loses the id — right. Each changed line was read in its context at the head:

  • build-progress.zod.ts lines 8, 27, 77, 84, 108, 123, 177: cloud#2172 becomes "the cloud build-progress card"; line 85: objectui#7388 block 2 becomes "block 2 of the objectui build-panel card". Every sentence keeps its verb and object — ruled the vocabulary in, owns WHERE in the loop a frame is emitted, the two carriers of the liveness watch, the emitter has not landed, owns the hop base. One reading nit, not a change: line 177 now reads "the cloud build-progress card owns and this card does not decide", two cards in one clause; it still parses to the same statement.
  • approval-service.ts:999: "exactly as [spec] Declare cancelRun and restoreConsumedSuspension on IAutomationService — #13953's ruled contract half, step (1), never filed #16495 declared IAutomationService.cancelRun / restoreConsumedSuspension" becomes "exactly as the automation-service contract declared …". The subject moves from the card that declared the two verbs to the contract that carries them; what the sentence tells a later author — copy that sibling's declaration form — is unchanged, and #15389 and decision batch #106 on the same lines stay.
  • notification.zod.ts:48 and :103, sharing.zod.ts:24 and :114: #5015 becomes "no-door retirement". The bodies the labels head say "had no authoring door anywhere" (notification, line 52) and "EmbedConfigSchema no door at all" (sharing, line 11), and "ADR-0049 enforce-or-remove, ruled REMOVE" stays on the next line, so the words name the decision the id stood for.
  • notification.zod.ts:94: [#4610] becomes [dual-source removal]. The note's own first sentence says "dual-source cleanup, ✅ spec 双源清账主单:基线 52 → 0(2026-08-03 收官)—— #4446 gate 落地后的偿还 worklist #4535 C3", and the correction block below it still names #4610 as the evidence it corrects, so the label describes the removal and the retirement's standing does not move.
  • content/docs/references/ai/build-progress.mdx: four body lines mirror the docblock; the front-matter description grows "(the cloud build-progress card, ruling A)" because packages/spec/scripts/lib/page-description.ts drops only citation-only parentheticals, so a parenthetical with words now stays. Deterministic generator output, confirmed by the green check:docs; noted in ③. ui/sharing.mdx:27 mirrors sharing.zod.ts:24.
  • Test strings: the two titles lose (#22110) and already state the decision (a flow text slot reads double-brace holes and refuses a single-brace token); the pin's ruling value and the why: ledger value keep their record ids (5979239990, 6070963704) and lose only the card number.

• Needles: same strength, and no place left matching the old text — right, measured on both refs. "Places" are the lines inside the exact text each test reads where the needle matches.

  pair	reader (verified in the test at the head)	old needle, places at base	new needle, places at head
  P1	build-progress.zod.ts, whole file (SOURCE)	cloud#2172: 8, 27, 77, 84, 108, 123, 177	"the cloud build-progress card": 8, 27, 77, 84, 108, 123, 177
  P2	the same	objectui#7388 block 2: 85	"block 2 of the objectui build-panel card": 85
  P3	approval-service.ts, the docblock slice above continueRestoredRun?(	#16495: 999 (the hit at 1083 is outside the slice at both refs)	"the automation-service contract": 999
  P4	ui/notification.zod.ts, indexOf plus the anti-vacuity regex at :134	// [#4610]: 94	// [dual-source removal]: 94
  P5	ui/notification.zod.ts and ui/sharing.zod.ts, each whole	#5015: 48, 103 and 24, 114	"no-door retirement": 48, 103 and 24, 114
  P6	the pin test's own ruling value at :322, matched by the regex at :417	#21704: the one value	"the five members the last typing stage stopped on": the one value

  • Each new needle has 0 hits at the base anywhere under packages/spec/src and content/docs, so none pre-existed to match vacuously. Each old needle has 0 hits at the head in its reader: cloud#2172, objectui#7388 block 2, [#4610] in ui/notification.zod.ts and #5015 in both ui/ modules are gone; #16495 remains in approval-service.ts only at :1083, outside the slice at base and head alike; [#4610] remains only in the ./system twin, which no test reads by that label.
  • scripts/ and scripts/adr-anchors/ name none of the old needles and none of the four reader files, so no gate read the old text.
  • The "every place moves" rule was applied consistently: cloud#2172 matched bare, so all seven occurrences were places; objectui#7388 block 2 matched only at 85, so the two other objectui#7388 mentions (:78, :107) are citations, not places, and rightly stay.

• Scope held — right. The 15 files are the claim's ten, the two class (e) titles, the amendment's metadata-form-zod-reconciliation.test.ts:497, ui/sharing.mdx (follows sharing.zod.ts:24 through gen:docs alone; check:docs would red without it) and the changeset. 0 governed paths; 80 changed lines; first line Part of #20749 with no closing keyword (both Part-of guards green). The three class (e) strings have no reader: a git grep -F of each old title and of the why: value's opening words at the head finds only the literal itself. With them gone, the only ids left in the packages/spec/src test strings are the four CSS hex colours stage 21's ACCEPT kept, which is the end state stages 20, 21 and 30 set for the card.

• Residual ids beside the moved text, read on purpose: objectui#7388 (:78, :107), cloud#1838, #15389, #106, #4535, #5781, #4610 (:106), #4583, #3950 are comment citations no test matches. The card body partitions code comments to packages/spec/src: 1,277 comment lines still cite 170 deleted tracker numbers (1,295 sites) — the staged remainder of ruling C+D on #19123, measured by PR #20226 #20234, so they are not this stage's share; raised once in ③.

② Semver level

• One changeset, '@objectstack/spec': patch, with Clause-②: no at a line start in both the changeset and the PR body — right. The diff publishes: three shipped .zod.ts sources and a .d.ts docblock change, so skip-changeset would be wrong (that label is for a diff that publishes nothing from any released package). Nothing authorable, exported or typed is added, removed or renamed, so patch is the level, neither (widening) nor (narrowing) applies, and no ADR-0087 disposition marker is owed. The changeset body carries no tracker number, names the four shipped texts and tells a consumer that matched the old words which new words to match. Check Changeset is green on both of its runs.
• Clause-②: no — upheld. No published surface grew or shrank.

③ Boundary flags

open_questions: the stage report declared one (the regrown why: ledger value at :497, outside the original claim). Answered A by the ACCEPT 6085085283, which amended the claim with that file, and executed as a766d439f: one span on one line, #22211 A becomes letter A, the ruling record 6070963704 stays, and no assertion reads the text — confirmed on the head. The patch report declares none.

deviations, stage report, each answered:

1. Reader places beyond the claim's "about" lines, inside admitted files (build-progress.zod.ts 77, 84, 108, 123, 177; sharing.zod.ts 114) — required, not merely accepted: leaving any would have left the old id standing at a place the old needle matched, so the new needle would match fewer places than the old one. Confirmed by the places table.
2. Two regex readers moved in lockstep (notification.test.ts:134, the pin at :417) — confirmed; both sit in claimed test files, and :417 is the reader stage 20's and stage 30's rulings named.
3. The build-progress page description keeps the words — confirmed from page-description.ts. Observation for the next edit of that docblock, not owed here: the lead parenthetical now ends the search snippet; stating the provenance in a second sentence would keep the snippet on the subject.
4. Base-side mutation legs on the one worktree, detached at the base; 5. the null P2 leg re-anchored before its run; 6. a refused bash -c loop re-run as literal commands; 7. stage 30's scratch caches overwritten (read caches, not evidence any record cites); 9. the stage instruments living in scratch, not the repo; 10. the cleanup order — process notes, not review matters.
5. Model-free commit trailers and the session-URL PR footer — confirmed on the PR body; neither the PR body nor the changeset carries a model identifier.

deviations, patch report, each answered:

1. The STALE TREE note (branch behind origin/main, gates derived on the branch tree) — accepted: the check-runs on this head ran on the merge ref and are the gate verdicts; main has moved further since, which the queue's rebuilt generation re-verifies.
2. A separate scratch directory; 3. the four observations unchanged; 4. trailers; 5. cleanup — not review matters.

out_of_scope_findings (four, all noted in the PR's Acceptance notes and none filed), each answered:

• Class (e) regrowth between stages — agreed, an observation about the denominator; the census at this head reads only the four kept colours, and nothing in this diff changes what arrives later.
• The build-progress liveness docblock's "zero readers" and carrier sentence may be stale after objectui's chat-panel change — agreed that it is documentation drift and that this stage was right to replace the id and leave the claim. Escalated to the seat, one line: the docblock's carrier sentence is a claim in shipped text with no named carrier; whether a docs-drift card is owed is the seat's call at close-out. Not a ground against this PR, which neither adds nor removes the claim.
• Comment citations with live ids in the four reader files — agreed, comments are packages/spec/src: 1,277 comment lines still cite 170 deleted tracker numbers (1,295 sites) — the staged remainder of ruling C+D on #19123, measured by PR #20226 #20234's share by the card body's partition. Escalated to the seat, one line: the dev report records packages/spec/src: 1,277 comment lines still cite 170 deleted tracker numbers (1,295 sites) — the staged remainder of ruling C+D on #19123, measured by PR #20226 #20234 as closed, so those residual ids now have no named carrier; the seat decides whether a successor card is owed. Not this PR's.
• The two-digit PD #12 / batch #58 labels — agreed, below the census id pattern and unchanged by this diff.

Not a flag, noted for the landing seat: the PR is a draft (ready first, arm second), and the ACCEPT orders the card to close when this stage lands, since its end state is then reached.

Implemented-by: claude/issue-20749-stage-31-needles
Reviewed-by: session_01DhTqaEHqPVSVnAkjG3jywn

VERDICT: PASS