docs(spec): anchor five dead tracker citations to commits, and measure the true clean-up scale

[objectstack-fleet]

Fixes #19123
Clause-②: no

What changed

Five lines in packages/spec/src/ui cited two tracker numbers GitHub no longer serves (#16714, #16715: REST 404, no 301, re-probed today against lit controls). They are rewritten on ruling C+D (comment 5749154545 on this card): each line now anchors to the commit in this repository's history that decided it, and says in its own words what that commit decided, so a reader never needs the link.

Line (base 4df101c38)	Was	Anchored to	What the line now says
app.zod.ts:457 (shipped as source)	(#16714)	commit 4cfc93b802	the filters docblock states no precedence order because objectui's hand-written mirror copied one and accepted a combination the schema refuses
app.zod.ts:504 (shipped as source)	EXPORTED (#16714)	commit 4cfc93b802	the guard is exported so a mirror chains the schema's own rule; no accept set moves
app-nav-target-exclusivity-export.test.ts:4	[#16714] header tag	commit 4cfc93b802, with PR #16862 beside it as a convenience link only	the three things that commit decided: export the guard, state no order in the filters docblock, move no accept set
app-nav-target-exclusivity-export.test.ts:243	test title (#16714 ruling)	the title now carries the lesson; the comment under it anchors to 4cfc93b802	"the export deliberately moved no accept set"
object-refinement-check-exports.test.ts:469	(#16715)	commit b37f0b1767, the diff that added the line	the declaration-count hardening closes a residual hole in the pin; no live shadowing existed

No ADR, ruling docblock or scripts/adr-anchors/ entry records either decision (objectNavTargetExclusivity in docs/adr and scripts/adr-anchors: 0 files), so both take the commit-sha rung of C. The two app.zod.ts lines are the two shipped-tarball lines the ruling puts under D first: they now carry no tracker number, only the lesson and a sha. The patch changeset for @objectstack/spec says a docblock citation changed and nothing else.

Prose, comments and one test title only. No assertion changes meaning and no accept set moves.

Premise, re-measured on 4df101c38

• The ruling's first D act, the refusal string in packages/spec/scripts/manifest-prescription.ts:78, has no live target. That file and check-react-blocks-declaration-parity.ts were retired by 28ce61260. git grep over the whole tree finds the three numbers on 5 lines, all in the table above, and #16697 on 0.
• #16715 has a durable in-repo object. Commit b37f0b1767 added the very line that cited it, and its message records the lesson: "a residual hole in the pin, not a live shadowing".

Census: the true clean-up scale (read-only, nothing swept)

Population. Every tracker-shaped token in packages/spec/src/** (1,594 tracked files) that names this repository: bare #N and objectstack-ai/objectstack#N, with N of 100 or more. Ordinals (batch #, summon #, Directive #, item # and similar) and 3- or 6-digit colour literals are excluded. Cross-repo tokens (objectui#, cloud#, ui#) are counted separately.

Instrument 1: the REST status of every number (a full census, not a sample)

GET /repos/objectstack-ai/objectstack/issues/N without following redirects, 2026-09-27T11:13Z to 11:16Z, plus 5 fully-qualified numbers at about 12:05Z.

status	numbers
probed	1,958 (every distinct in-repo number)
200	1,786
404	172 (8.8%)
301	0

Controls. They were re-probed at the start, after every 250 probes and at the end, 9 checkpoints in all. The lit controls #16862, #16847 and #17698 answered 200 at 9 of 9. The dead controls #16714, #16715 and #16697 answered 404 at 9 of 9. So the channel stayed up for the whole run, and the 404s are real readings.

The dead rate depends on the number's era (bare numbers):

numbers	dead
below 5000	0%
6000s	21.2%
7000s	0%
8000s	14.0%
9000s	10.6%
10000s	40.0%
11000s	26.7%
12000s	18.5%
13000s	16.4%
14000s	22.5%
15000s	0%
16000s	14.3%
17000s	10.5%
18000s	7.7%
19000s	3.5%

Contamination bound. Five numbers are cited only as sibling-repo PRs, written bare (for example objectui PR followed by the number). One of them, 8887, is among the 404s, and 8758 is cited both ways. So the in-repo dead count is 170 to 172 out of 1,953 to 1,958.

Cross-repo tokens. objectui (183 distinct) and cloud (36) are NOT MEASURED: this session's REST gate answers 403 for both repositories. hotcrm answered 200 for 7 of 7 (lit control hotcrm#1961: 200).

Instrument 2: who owns the squash commit

GET /repos/objectstack-ai/objectstack/commits/SHA/pulls for every dead number that appears as a squash suffix in origin/main history (8,715 commits, back to 2026-06-22). 11 of the 172 do, and all 11 answer []: the PR object is gone. Controls: b37f0b1767 answers [16847] and 1124c45641 answers [19107] (lit); 7936b29ce5, the card's known-dead #16697, answers []. The other 161 dead numbers carry no squash suffix. They were cards, not PRs, and this instrument does not reach them.

Independent cross-check: check-issue-citations --census

I ran the repository's own citation gate in census mode at about 12:02Z. The board frontier is 20,223 and the board size is 18,050, so 2,173 minted numbers (10.7%) are holes. Tree-wide there are 2,779 unresolvable sites, 664 of them in packages/spec/src.

On the gate's surface (non-test comment prose), the two instruments agree on 661 of those 664 sites. The 3 sites only the gate found use the objectstack-ai/objectstack#N spelling, which my first extraction pass missed; I probed them afterwards and they are included above. My 297 extra comment sites break down as 271 in test files (the gate's declared deferred surface) and 26 re-charter #13135 sites that the gate reads as an ordinal.

What remains after this PR

On this head, packages/spec/src/** still has 1,295 dead-cited sites on 1,277 lines in 327 files, citing 170 numbers. On the base it was 1,300 sites, 1,282 lines, 330 files and 172 numbers.

class	sites	lines	files
non-test comment	691	683	180
non-test string or code (text an author can be shown)	186	184	48
test comment	271	266	103
test string or code (titles, fixtures)	147	144	54

By area: migrations 438, data 239, api 153, ui 107, kernel 105, contracts 81, system 38, conversions 24, shared 20, other 90.

Load-bearing share. I read a seeded random sample of 40 dead-cited lines. 37 of the 40 cite a ruling, a deliberate tolerance or exclusion, a retirement, or a cautionary measurement, which is the dispatch's definition of load-bearing. The other 3 are incidental: a bare Refs, a label, a process note. Applied to 1,277 lines, that is roughly 1,100 to 1,200 load-bearing lines, a sweep the ruling keeps out of a single PR. None of them is rewritten here.

The runtime-text slice is the D half. os migrate meta prints each migration entry's replacement, why: (reason) and verify: (acceptanceCriteria) to the author (packages/cli/src/commands/migrate/meta.ts:523-525). Across packages/spec/src/migrations/**, those strings hold 2,060 tracker-id sites (478 distinct numbers; 178 sites are dead). pnpm check:doc-authoring is green over all of them, for two reasons: its Rule 3 does not recognise those property positions, and its cross-package leg excludes packages/spec/src. I report this to the seat as a finding and do not fix it here.

Every dead number still cited, as number and sites/files, in numeric order. Any line is recoverable with git grep -n on the number under packages/spec/src.

• #5970 6/3 · #6037 11/5 · #6048 8/3 · #6072 1/1 · #6075 2/2 · #6083 10/7 · #6085 2/1 · #6111 3/3
• #6148 10/6 · #6206 14/7 · #6216 2/2 · #6238 1/1 · #6239 10/6 · #6259 5/2 · #6276 13/2 · #6279 2/2
• #6287 18/2 · #6300 2/1 · #6306 1/1 · #6345 32/11 · #6361 20/11 · #6362 9/5 · #6363 12/6 · #6483 6/1
• #6511 7/3 · #6523 11/5 · #6527 1/1 · #6571 10/2 · #6573 2/2 · #6604 3/2 · #6605 3/1 · #6640 2/2
• #6704 6/3 · #6723 1/1 · #6725 5/2 · #6745 2/1 · #8460 2/1 · #8495 59/27 · #8656 1/1 · #8696 34/13
• #8714 4/2 · #8715 47/24 · #8758 9/6 · #8772 11/5 · #8778 1/1 · #8794 5/3 · #8836 4/3 · #8873 10/6
• #8874 1/1 · #8876 11/7 · #8885 2/1 · #8887 2/2 · #9040 20/6 · #9041 28/5 · #9740 3/1 · #9741 23/2
• #9933 9/2 · #9934 4/4 · #9972 4/2 · #10165 7/2 · #10194 12/4 · #10238 1/1 · #10264 1/1 · #10274 7/3
• #10329 10/4 · #10330 3/1 · #10338 6/3 · #10347 5/2 · #10485 43/23 · #10527 3/2 · #10627 12/7 · #10724 47/20
• #10726 15/9 · #10812 6/5 · #10926 11/6 · #11006 15/3 · #11065 7/3 · #11071 1/1 · #11166 2/1 · #11195 5/3
• #11215 1/1 · #11284 10/4 · #11328 2/2 · #11330 2/1 · #11331 2/1 · #11332 19/7 · #11333 3/3 · #11350 9/5
• #11408 3/2 · #11453 1/1 · #11504 5/3 · #11507 8/4 · #11513 2/2 · #11658 6/3 · #11693 2/2 · #11703 3/2
• #11741 9/4 · #11753 6/2 · #11757 4/2 · #11846 30/10 · #11858 1/1 · #11859 2/1 · #12010 5/2 · #12144 2/1
• #12165 1/1 · #12176 2/1 · #12194 4/4 · #12245 2/1 · #12248 17/4 · #12380 4/2 · #12493 6/2 · #12868 9/6
• #12950 3/2 · #12961 10/2 · #13031 3/2 · #13109 1/1 · #13135 70/25 · #13156 4/4 · #13195 3/1 · #13197 1/1
• #13218 7/2 · #13608 5/2 · #13644 3/2 · #13670 1/1 · #13672 3/2 · #13681 1/1 · #14143 2/2 · #14162 1/1
• #14192 5/2 · #14312 1/1 · #14365 4/3 · #14369 24/13 · #14419 2/2 · #14426 2/2 · #14474 1/1 · #14526 2/2
• #14662 4/2 · #14676 22/10 · #14686 7/2 · #14691 50/16 · #14722 4/2 · #14723 1/1 · #14748 2/2 · #14902 2/2
• #14919 2/2 · #14996 1/1 · #16126 2/2 · #16559 2/1 · #16626 3/3 · #16649 12/2 · #16659 5/4 · #16685 4/2
• #16786 3/2 · #16864 2/2 · #16867 3/2 · #17014 3/2 · #17058 1/1 · #17081 1/1 · #17124 3/3 · #17147 5/2
• #17286 1/1 · #17348 1/1 · #17590 1/1 · #17987 13/2 · #18003 1/1 · #18012 10/5 · #18177 7/2 · #18335 1/1
• #19307 1/1 · #19377 8/4

Gates (head 98edadd03)

• Derived gate families. node scripts/pm/dispatch-gates.mjs --ran reports 83 derived, 79 run with exit 0, 4 NOT MEASURED and 0 unrun. The four exited 3 (PREREQUISITE NOT MET): check:doc-formula-expressions, check:dual-build-cjs-loads, check:lean-entry-closure and check:type-check-debt. Each needs other packages' dist/, and both attempts to take the shared verify lock for that build ended in queue-timeout (99). This is a declared narrowing: the diff changes comment and test-title text in one package and nothing those four gates read, and CI's Lint & Repo Gates runs them.
• Roster gates. The six artifact-roster gates whose roster sits under a changed path (check-changeset-fixed, check:meta-url-spelling, check:spec-changes, check:authz-resolver, check:error-code-casing, check:filter-alias-parity) all exit 0.
• Build, tests, typecheck. pnpm --filter @objectstack/spec build exits 0. The two edited test files pass: 2 files, 141 tests. pnpm --filter @objectstack/spec typecheck exits 0.
• Citation verdict. node scripts/check-issue-citations.mjs --base 4df101c38 (the diff-scoped verdict) exits 0. The change adds one citation, #16862, and it answers 200.

Acceptance notes

• This change adds one tracker number: PR feat(spec): export objectNavTargetExclusivity; state no precedence order in the filters docblock #16862, once, in a test docblock, as a convenience link beside the sha it cites. The ruling allows that, and nothing else in the change cites a number.
• Where this departs from the dispatch. The dispatch said to rewrite any load-bearing dead-cited line found in packages/spec/src beyond the five, in this PR. At the measured 1,277 lines, about 92% of them load-bearing, that is the sweep the ruling keeps out of one PR, so this PR measures and lists them and rewrites none.
• D's first named act, the refusal string, was retired upstream by 28ce61260. The live D population is the migration-entry text above.
• Instrument notes, reconciled above and not filed: the gate's ordinal list reads re-charter #N as a non-citation, which hides 26 real citations of #13135. My own first extraction pass missed the owner/repo#N spelling (3 sites).

---

Generated by Claude Code

[github-actions]

📓 Docs Drift Check

1 anchor(s) derived from 1 changed package(s); no hand-written page names any of them, so this run has nothing to list — not a clean bill of health. This check sees only pages that NAME a derived anchor: one that documents this change in prose, or enumerates it in an authoring dialect, names none and stays invisible to it on every run.

What this run could not see

• 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 — 136 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 805af4f290565955d6e6b56ee46fed45f721d955 → packageMentionDocs.

Which tree this was computed on

This run read content/docs from 7816169646105a021c7ac14d03a1773d1c097994 — the merge of head 98edadd03c06f76737180e09c613cd69024305fc into base 805af4f290565955d6e6b56ee46fed45f721d955, 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 7816169646105a021c7ac14d03a1773d1c097994 && git checkout 7816169646105a021c7ac14d03a1773d1c097994
# afterwards, rebuild it from the two parents, which stay fetchable
git fetch origin 805af4f290565955d6e6b56ee46fed45f721d955 98edadd03c06f76737180e09c613cd69024305fc && git checkout -B drift-repro 805af4f290565955d6e6b56ee46fed45f721d955 && git merge --no-ff 98edadd03c06f76737180e09c613cd69024305fc

node scripts/docs-audit/affected-docs.mjs --json 805af4f290565955d6e6b56ee46fed45f721d955

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

[objectstack-fleet]

Contract review

Served-tier: CONTRACT_REVIEW_TIER
Head-sha: 98edadd03c06f76737180e09c613cd69024305fc

_{Read: card #19123 body and all 6 comments (ruling 5749154545, pointer 5807032824, dispatch 5855287758, dev report 5855787862), PR #20226 object/body/files/commits/check-runs, the full diff against merge-base 4df101c, the four files at the head, anchor commits 4cfc93b and b37f0b1 (stat, message, relevant hunks), retirement commit 28ce612, AGENTS.md (citation rule, changeset step 3, Touched packages/spec, Prime Directive 14), the headers of check-changeset-no-major / check-empty-changeset / check-changeset-fixed and check-issue-citations (surfaces, deferred table, cause table). Ran: a TypeScript-scanner token comparison of app.zod.ts base vs head with trivia dropped, an independent comment-region state machine over the same pair, git grep sweeps at base and head, REST probes of 17 issue numbers and 3 commit-to-PR associations, census arithmetic, 4 site-count reproductions, a bounded CI poll to convergence. NOT MEASURED: a local run of the two test files (the shared checkout has no node_modules; CI Test Core green stands in), the 1,958-probe census totals and era table, the check-issue-citations --census totals and the 40-line sample (spot-checked, not re-run), and the four derived gates the dev declared NOT MEASURED (CI ran them).}

① Derived judgments

(a) No accept set or runtime string moves. app.zod.ts base vs head through the TypeScript scanner with comment and whitespace trivia dropped: 3974 tokens each, identical token for token. Independent state machine: the 10 changed lines all lie inside /** ... */ blocks, none contains a comment terminator, and the code text is identical once comments are stripped. Every .describe() string, the two ctx.addIssue messages and the export function objectNavTargetExclusivity declaration are byte-identical. No runtime string changes.

(b) No assertion changes meaning. app-nav-target-exclusivity-export.test.ts changes a header docblock (lines 3 to 9), one it title (line 247: the old suffix "(#16714 ruling)" becomes "deliberately moved no accept set") and the comment under it; the two expect calls in that test are untouched. object-refinement-check-exports.test.ts changes three comment lines (468 to 470) above an unchanged declarations helper. The old title text appears nowhere at the head; at the base its only other occurrence is a prose line in the same header, not a pin. No snapshot, no gate keyed on it titles, and no gate requiring a [#N] test-header tag (the [#id] regex in check-doc-anchors / check-docs-single-h1 is a Markdown heading custom id on the docs surface). Leg 5's docblock extraction anchors on recordId: (head line 430), which precedes both rewritten docblocks (441 to 461 and 496 to 515), and leg 3 matches code lines only, so no assertion reads the changed prose. CI Test Core: success.

(c) Anchors are faithful and on the right rung. 4cfc93b802 is the squash of PR #16862 (commits/4cfc93b802/pulls answers [16862]); its app.zod.ts hunk turns the const arrow into export function objectNavTargetExclusivity, deletes "Precedence: recordId → filters → viewName" and inserts "There is deliberately NO precedence order stated here", and its changeset states "No schema's accept set moves" and that objectui's mirror "copied that first half" of the docblock. That is exactly what the rewritten filters docblock, the EXPORTED (commit 4cfc93b802) line, the test header's three-item list and the it-comment attribute to it. b37f0b1767 is the squash of PR #16847 (association answers [16847]); its hunk adds the declarations helper and the very comment now rewritten, and its message reads "four names, one declaration each — a residual hole in the pin, not a live shadowing", which is what the new comment says. Ruling C's first rung is empty: objectNavTargetExclusivity / target-exclusivity in docs/adr, scripts/adr-anchors, content and .claude: 0 files; the numbers 16862, 16847, 16714, 16715 in docs, scripts/adr-anchors, content and .claude: 0 hits; docs/ has no rulings directory; the guard docblock's ADR-0053 mention is the guard's design philosophy, not the export decision. The commit-sha rung is therefore correct for both. The one PR number, #16862, sits once in a test header expressly labelled a convenience link beside the sha; the shipped lines and the it-comment cite the sha alone. No misattribution, no PR-as-citation.

(d) Ruling D on the two shipped lines. Head 457 to 460: "objectui's hand-written mirror did exactly that, which is why commit 4cfc93b took the ordering out of this docblock" and head 506: "EXPORTED (commit 4cfc93b), one function per refinement, the same posture" carry no tracker number and each states the decision in words (the ordering was removed because a mirror copied it and accepted what the schema refuses; the guard is exported so a mirror chains the schema's own rule), so a reader needs no link. Caveat, non-blocking: the same sentence continues on head line 507, re-wrapped by this diff, with "the check* exports of #16489", a live issue (REST 200) outside the card's measured set; see ③.

(e) The one added tracker number, PR #16862, answers REST 200 (a closed, merged pull request). check-issue-citations lists packages/**/*.test.ts in DEFERRED_SURFACES and applies it as an exclusion in surfaceFor, so the header is not judged; on a judged surface a number that resolves as a pull request classifies resolves-as-pull-request, which is not in FINDING_CAUSES. On the judged surface the diff-scoped run sees #16489 on the re-wrapped app.zod.ts line, which resolves. CI Lint & Repo Gates, which runs pnpm check:issue-citations && node scripts/check-issue-citations.mjs: success.

(f) Census. Method: REST issues/N without following redirects with lit and dead controls at nine checkpoints, and commits/SHA/pulls for squash suffixes; both instruments reproduce today (16862, 16847, 17698: 200; 16714, 16715, 16697: 404; 4cfc93b → [16862], b37f0b1 → [16847], 7936b29 → []). Arithmetic: 1,786 + 172 = 1,958; 172/1,958 = 8.78%; 11 + 161 = 172; base to head deltas (1,300→1,295 sites, 1,282→1,277 lines, 330→327 files, 172→170 numbers) equal the 5 sites, 5 lines, 3 files and 2 numbers the diff removes; the class table sums to 1,295 sites / 1,277 lines; the area table sums to 1,295; the per-number list has 170 unique sorted entries summing to 1,295 sites and omits 16714 and 16715. Spot checks: eight sampled dead numbers (5970, 13135, 19377, 8887, 8758, 6037, 17987, 16575) all 404; site/file counts for 19377 (8/4), 5970 (6/3), 6037 (11/5) and 13135 (70/25) reproduce exactly by git grep -o -w at the head. Tree claims: 1,594 tracked files under packages/spec/src at the head; the three numbers on exactly 5 lines at the base and #16697 on 0; manifest-prescription.ts and check-react-blocks-declaration-parity.ts deleted by 28ce612 and absent at the head; origin/main reaches back to 2026-06-22 (8,720 commits now against the dev's 8,715). The body says "None of them is rewritten here" and the diff touches only the five lines, so nothing claims a sweep the diff does not do.

② Semver level

patch for @objectstack/spec with Clause-②: no is the right level. check-changeset-no-major's header forbids major (none here) and binds a minimum only on a PR that declares clause ② (this one declares no, readably). check-empty-changeset's header rejects an empty frontmatter (this one names the package and bump) and a modified foreign changeset (this file is added by the PR). AGENTS.md step 3: a changeset for anything that publishes, skip-changeset only when nothing published moves; src/**/*.zod.ts is in the package's files, so the docblock ships and a changeset is owed, and with no export, accept set, .describe() or error message moving, nothing above patch is. The changeset body carries no tracker number. The "Touched packages/spec" table names check:docs for a TSDoc edit; no committed artifact carries the field docblock text (the old phrase appears nowhere but app.zod.ts at the head), and the TypeScript Type Check job that runs the generated-artifact gates is green. Check Changeset: success.

③ Boundary flags

Blocking: none
Non-blocking: (1) No sweep beyond the five lines: the ruling says the true scale goes in the PR body and is not swept in one PR; the dispatch's "any further dead-number line the census finds" clause meets a measured 1,277 lines, about 92% load-bearing, and the departure is declared in Acceptance notes. Consistent with the ruling. (2) Migration-entry author-facing strings (2,060 sites, 178 dead) reported as an out-of-scope gate-reach finding beside #19124 rather than fixed: a different area and a different gate; not this card's file surface. (3) Four derived gates NOT MEASURED locally with exit 3 (check:doc-formula-expressions, check:dual-build-cjs-loads, check:lean-entry-closure, check:type-check-debt): CI's Lint & Repo Gates and Type Check · debt ledger ran them, both success. (4) #16489 remains on the re-wrapped shipped line 507 and in the test header line 11: live today, outside the card's dead set, but the same sentence the PR rewrote still ends in a tracker number; a candidate for the staged C sweep. (5) The body's "the change adds one citation, #16862" describes the diff, not the gate's judged surface, where the added citation is #16489 and the test header is deferred; both resolve, so nothing turns on it. (6) A local run of the two edited test files is NOT MEASURED here; CI Test Core covers it.

CI at this head: 35 check-runs on 98edadd, converged 12:40Z: 32 success, 3 skipped (Build Docs, Console Pin Gate, Packed-tarball smoke (opt-in)), 0 failed, 0 pending. All seven required contexts from the main-branch ruleset are success: TypeScript Type Check, Test Core, Dogfood Regression Gate, Build Core, Temporal Conformance (live PG + MySQL), Lint & Repo Gates, Governed Surface Queue Guard. mergeable_state clean, PR still draft. Closing keywords: the body carries Fixes #19123 exactly once and no other closing keyword; "The card this PR closes must claim this branch" and "Part-of PR must not also close its card" are success.

Implemented-by: claude/issue-19123-citation-anchors
Reviewed-by: session_01CiCTczDo7tGhafXjf61dUJ

VERDICT: PASS