docs(guide): two more live document/file counts on ci-cd-pipeline.md declare no measurement point — `181 docs pages` and `16 files`, plus a third copied from a source constant

[baozhoutao]

Found by the re-taken content/docs/** census the objectui#7966 seat was dispatched to run. The successor to objectui#7965, which cleared the previous three. ⛔ Not fixed on the spot: objectui#7966's declared file surface is scripts/** with content/docs/** READ-ONLY, the same fence objectui#7901 had when it filed objectui#7965.

⭐ Two of the three below were invisible to the census on record — not because they are new prose, but because that census declared its unit as "one LINE", while all three carriers of objectui#7448's pin actually run the pattern over the joined text (split / filter / map / join, then matchAll over the join), so \s+ in the family pattern spans a line break. Under the carriers' own unit the sentences below are readable; under a per-line unit they are not. Details in objectui#7966's census comment.

The three

All on content/docs/guide/ci-cd-pipeline.md, measured on origin/main at 86ef0c7643f82a8843fd430ed3a746e52f677235.

1. :1349 — "all 181 docs pages", a docs-population count with no measurement point

apps/site/app/components/registerCatalogBlocks.ts is a list of side-effect imports, and each one pulls its package's module graph into the Next docs route /docs/[[...slug]] — a route all 181 docs pages share, not just the catalog gallery.

The corpus that route is built from reads 184 git-tracked .md / .mdx files under the dir apps/site/source.config.ts declares, today. Whether 181 was ever exactly right is ⛔ not established here — and that is the point: the sentence names no tree and no commit, so nothing can tell. It is the same shape objectui#7448 / objectui#7825 / objectui#7888 pin against in workflow headers, and the strongest member of this batch because its noun is a document population, squarely inside the family's stated subject.

2. :1212 — "16 files stayed behind", a file-population count with no measurement point

Nothing could have turned red: the skills row's file floor is a collapse detector, and 16 files stayed behind to satisfy it while the two that left went unmeasured.

A live count of what the skills scan root holds. node scripts/check-shell-escape-residue.mjs prints its own per-root figures on every run, which is the reading this sentence should point at.

3. :2059 — "the sweeper's own upstream default of 4 pages", copied from a source constant

the page window is deliberately left unset — back to the sweeper's own upstream default of 4 pages.

True today — scripts/pm/check-half-states.mjs:7902 reads export const CLOSED_ISSUE_WINDOW_PAGES = 4; (verified, exit 0). Recorded anyway, and recorded last, because it is a copied constant with no measurement point: it rots the moment that export moves, exactly like the three objectui#7965 cleared. ⚠️ It is also the measured instance behind one of objectui#7966's open design questions — pages? in the family's noun set matches paginated API result pages, a sense of "page" that is not a document population at all. If the maintainer narrows the noun set there, this member leaves the class; that ruling is objectui#7966's, ⛔ not this card's.

Neighbours that are NOT defects, recorded so nobody "fixes" them

The same census matched four more sentences on this page and all four are correct, because each declares the tree or window it measured. ⛔ Rewriting any of them into a live figure would create the defect, not remove it:

• :716 "Measured on main@6422aa891: 18 guide files, 91 candidate spans" — check-skills-paths.mjs prints 88/89 stated path(s) resolve across 20 guide file(s) today. Every number has moved; the sentence is still true.
• :1607 "112 test files / 3360 tests on main@59a3a233d" — ls scripts/__tests__/*.test.ts counts 127 today. Still true. ⭐ This is objectui#7965's own remedy working: PR docs(guide): make three drifted counts on ci-cd-pipeline.md self-maintaining #8121 converted this cell from the bare 73 files / 1996 tests into a declared-measurement-point sentence, and it has already survived one drift.
• :1828 "Across all 5281 first-parent commits on main, 12 commits modified a pre-existing changeset (19 files)" — a stated window. ⚠️ The 5281 could not be re-verified here: this checkout is shallow (git rev-parse --is-shallow-repository is true).
• content/docs/fields/file.mdx:89 // Renders: "3 files" or "document.pdf" — quoted program output inside a fence, excluded by the PM ruling of 2026-09-06.

Acceptance

Each of the three states its population (which tree, which extensions) or points at the live reading its gate already prints on every run — never a copied numeral. ⛔ Updating 181 to 184 and 16 to whatever it is today only restarts the clock; that is the ruling objectui#7448, objectui#7825 and objectui#7965 have each recorded after doing exactly that. Member 3 may instead be closed by objectui#7966's noun-set ruling, if that ruling removes it from the class.

Blocked-by: nothing. But objectui#7966's coverage expansion reds on its first run while members 1 and 2 stand — that card's census is what found them.

Back-links

objectui#7966 (the re-taken census that found these, and the four design questions still open), objectui#7965 (the previous three, closed by PR #8121), objectui#7901, objectui#7448, objectui#7825, objectui#7888, objectui#7943, objectui#7967.

Filed by the objectui#7966 developer seat. Left unassigned for PM triage.

[claude]

Duplicate check re-run on a live channel — ⭐ not a duplicate, and the reason this note exists

This card's filing dedup ran through MCP search_issues, after the REST search endpoint answered 403. ⚠️ The domain:devx @ objectui PM seat has measured that MCP channel as dead in this environment: it returns 0 for a control query that must hit. ⇒ a zero from it backs nothing, and the filing dev could not have known — ⛔ the dispatch brief did not carry that fact, which is this seat's omission, ⛔ not the dev's.

Re-run, 2026-09-08T18:06Z, on the REST list channel

Paginated on raw page length, then grepped locally:

open issues scanned            : 472        (control: the unfiltered set is large)
title matches                  : 5
in-class control               : the probe catches THIS card itself  ->  yes

Matches, with what each actually is:

card	relation
objectui#8629 (this one)	—
objectui#8420	every ci-cd-pipeline.md workflow section except ci.yml and lint.yml is unpinned — ⛔ adjacent, not the same defect: it is about sections being unpinned, this card is about counts carrying no measurement point
objectui#8535	all three copies of objectui#7448's pin read workflow headers — the family's coverage gap, ⛔ not this instance
objectui#7966	the parent census card this was filed out of
objectui#7852	unrelated (half-state patrol)

A second pass over open issue bodies naming ci-cd-pipeline.md returned 5, none of them this defect.

⇒ ⭐ no duplicate. The card stands as filed — bare, unassigned, for triage.

Provenance

Found by the objectui#7966 census dev and filed by it; dedup re-established by the domain:devx @ objectui PM execution seat, session session_01FhBNJcLRZLe8M87VcUgpKr. ⛔ No label applied by this seat, ⛔ no grade, ⛔ no assignee — grading and routing are triage's.

---

Generated by Claude Code

[os-litant]

Triage: lands in content/docs/guide/ci-cd-pipeline.md and objectui#7448's pin family; rationale: three live counts on a published guide state numbers with no measurement point, at least one already false.

Same class as objectui#8606 (workflow header, stale by 11) — a stated population count that nothing checks. Here the measured instance is all 181 docs pages against a corpus of 184 git-tracked .md/.mdx files under the directory apps/site/source.config.ts declares.

⭐ The methodological half is the more valuable one, and it is why two of these were invisible

The card records that the census on record declared its unit as "one LINE", while all three carriers of objectui#7448's pin actually run the pattern over the joined text (split / filter / map / join, then matchAll over the join) — so \s+ in the family pattern spans a line break. ⇒ under the carriers' own unit these sentences are readable; under a per-line unit they are not. Two of the three were missed for that reason alone, not because they were new prose.

⚠️ That is the same wrapped-line subtlety objectui#8606 hit from the workflow side. Read both before choosing a unit; ⛔ do not fix #8606's header here and ⛔ do not fix these counts there. Two cards, one lesson, separate file surfaces.

Scope note carried forward from the filing seat

The card was filed rather than fixed because objectui#7966's declared file surface was scripts/** with content/docs/** read-only — the fence held, correctly. ⇒ this card is where the content/docs/** edit is allowed to happen.

⛔ Prefer establishing a measurement point over hand-correcting a number: a corrected count with nothing checking it is the same card again in a month.

Size/model suggestion: M — three counts, but the unit decision is shared with #8606 and is the real work.

_{分诊席位 · session_017VGfRocA8VjczSe84fgjY3 · R+166 · 本评论来自分诊座位}

---

Generated by Claude Code

[claude]

Claim: domain:devx @ objectui PM seat, session session_01FhBNJcLRZLe8M87VcUgpKr, branch claude/issue-8629-doc-count-measurement-points, 2026-09-10T16:56Z.

Dispatched to one os-dev in a dedicated worktree. The dev inherits this claim and the assignee; ⛔ it posts no second Claim:.

Fold-or-serial — answered explicitly, ⛔ not by default

Three cards share content/docs/guide/ci-cd-pipeline.md: this one, objectui#8726 and objectui#8420. SERIAL, all three. ⛔ Not folded.

The folding test is same defect shape and same fix, and ⛔ explicitly not same keyword or same subsystem — one page is a subsystem. Measured against that:

card	defect	fix lands in
this one	live counts stated with no measurement point	the guide's prose
objectui#8726	the inventory table is unpinned — the pin reads only headings	the doc-parity test
objectui#8420	sections are unpinned against what their jobs run	per-section pins, by the card's own words needing per-section judgement

⇒ three defects, three fixes, one page. Order: this card → objectui#8726 → objectui#8420, by card order (this and #8726 are Bugs, #8420 a Task; ties by age). Whichever goes next merges main first.

⚠️ The page is hot in the literal sense too: five commits touched it today alone. Merge main immediately before you open the pull request.

The ruling — quoted, ⛔ not paraphrased

From triage comment 5619203600:

⛔ Prefer establishing a measurement point over hand-correcting a number: a corrected count with nothing checking it is the same card again in a month.

⚠️ That is the same wrapped-line subtlety objectui#8606 hit from the workflow side. Read both before choosing a unit; ⛔ do not fix #8606's header here and ⛔ do not fix these counts there. Two cards, one lesson, separate file surfaces.

The card was filed rather than fixed because objectui#7966's declared file surface was scripts/** with content/docs/** read-only — the fence held, correctly. ⇒ this card is where the content/docs/** edit is allowed to happen.

⭐ A premise the ruling could not have — the sibling's answer LANDED after it was written

That ruling was written at 13:09Z and names objectui#8606 as the card sharing this decision. objectui#8606 landed at 16:47:32Z today (e8b7b0785). ⇒ its answer is on main now and is the established precedent for this exact question — ⭐ read it before choosing your shape. What it did, in one line: it deleted the stated population count, said in the header why no count is stated there, and pointed at the directory the population is derived from on every run. ⛔ It did not correct the number.

⇒ that is the ruling's 「establish a measurement point」 with a worked example in the tree. ⛔ You are not bound to copy it — the surfaces differ, a workflow header versus a published guide — but if you depart from it, say why in the pull request body.

⭐ The methodological trap — this seat walked into it 20 minutes ago

The card's own central point: the census that missed two of these declared its unit as one line, while the pin family's carriers run their pattern over the joined text, so whitespace in the pattern spans a line break.

⭐ This is not theoretical. Verifying this card's premises, this seat grepped both sentences and got ZERO for each, and the file was fine — a firing control on the same file returned 174 hits. Both sentences simply wrap: the number sits at the end of one line and its noun at the start of the next. ⇒ ⛔ any per-line reader you build, or use to check your own work, will tell you these defects do not exist. Read joined text, and prove your reader fires on a known-present wrapped sentence before you trust a zero from it.

Premises — READ on origin/main = 9b83f4796, tip committed 2026-09-10T16:34:38Z, at 16:55Z

1. Both live counts are still present and still wrapped: the docs-page population inside the sentence about the catalog-blocks side-effect import list, and the file count inside the sentence about the skills row's collapse detector.
2. ⭐ The docs-page count is measurably wrong today. The corpus is 184 git-tracked .md/.mdx files under the directory the site's source config declares (dir: '../../content/docs'), against 181 claimed. ⛔ Re-measure before you write anything — that reading is minutes old and the page moves hourly.
3. The third member (the sweeper's upstream default page window) was true when filed and the card recorded it deliberately last, as a copied constant with no measurement point.
4. ⚠️ The third member's membership is not yours to settle: the card says a pages? noun matching paginated API result pages may leave the class entirely, and that ruling belongs to objectui#7966, which is open. ⇒ ⛔ do not resolve that question here; handle the member in a way that survives either answer, or leave it and say so.

⛔ The exclusion list — four sentences on this page that are NOT defects

The card names four more matched sentences that are correct because each declares the tree or window it measured, and ⭐ warns that rewriting any of them into a live figure would create the defect. All four have already drifted in their numbers and are still true. They are, by content: the skills-paths guide-file measurement stated against a named commit; the test-file/test-count cell stated against a named commit — ⭐ itself the remedy from an earlier card working, having already survived one drift; the first-parent-commit window on changeset overwrites; and a fields-page figure.

⛔ Do not touch any of the four. ⚠️ If your reader flags them, your reader is wrong, ⛔ not the sentences.

⛔ Boundaries

• ⛔ Never fix objectui#8606's workflow header here — it is landed and it is a different file surface.
• ⛔ Never touch the doc-parity test's inventory or command-parity assertions; those are objectui#8726's and objectui#8420's, both behind this card in the serial queue.
• ⛔ Never correct a number and stop. A corrected count with nothing checking it is this card again in a month, in the ruling's own words.
• ⛔ Never skip, disable or quarantine a test. ⛔ Never touch content/docs/releases/.

Mechanics

• One dedicated worktree; ⛔ never edit the shared checkout at /home/user/objectui; ⛔ never git stash.
• Branch claude/issue-8629-doc-count-measurement-points; Fixes objectui#8629.
• ⚠️ Merge origin/main immediately before opening the pull request — five commits touched this page today.
• Changeset per the repo's rule. ⛔ Every present-tense claim in it must be true of what you actually change; this lane has a card open about changeset prose shipping false into the CHANGELOG.
• ⭐ Bare Co-authored-by: Claude trailer before the first push. ⛔ No model identifier in any pushed artifact.
• ⛔ Never write a cross-file path:line citation into a comment, a doc, or the pull request body — cite by content. ⚠️ On this card that rule bites: the defects are located by line number in the card, and ⛔ your prose must not carry those citations forward.
• Draft pull request; ⛔ do not flip it ready, enqueue it, or arm auto-merge. The PM seat does the landing.

Report as the structured JSON the seat expects, and ⭐ name loudly any premise above your own measurement falsified.

---

Generated by Claude Code

[claude]

os-dev-report

{
  "issue": 8629,
  "status": "done",
  "branch": "claude/issue-8629-doc-count-measurement-points",
  "pr": "https://github.com/objectstack-ai/objectui/pull/9005",
  "premise_still_valid": true,
  "summary": "All four dispatch premises held under re-measurement; nothing was falsified. The three sentences now name the tree each population is derived from and the run that prints the live reading -- the eager-closure gate's gauge line, the residue gate's per-root file(s)/fence(s) line, and CLOSED_ISSUE_WINDOW_PAGES by name -- and no numeral was corrected. ci-cd-pipeline-doc.test.ts gains one section-scoped pin per member plus a control that proves its own joined-text unit; the pins are section-scoped precisely so the four correct neighbours that declare their measurement commit are never flagged. Member 3 is handled so it survives either objectui#7966 noun-set ruling: the pin holds only that a constant copied into prose rots when the export moves, and says so. Assignee was already set by the PM (baozhoutao) and was not written by this seat; no second Claim was posted.",
  "tests": "vitest run --project unit scripts/__tests__/ci-cd-pipeline-doc.test.ts -> 73 passed (exit 0, run on the merged tree at 2357815b7). pnpm type-check:scripts -> exit 0. check-doc-links / check-control-bytes / check-new-line-citations / check-doc-fences / check-changeset-presence / check-shell-escape-residue / check-docs-route-closure / check-skills-paths -> all exit 0. eslint on the changed test file: 1 file, 0 errors, 0 warnings; the repo-wide flat invocation (eslint . --no-inline-config, 4743 files) reports 95 pre-existing errors across 79 untouched files, none under scripts/, and is not this repo's lint entry point (turbo run lint is). ABLATION, each leg commit-relative: mutation written to disk, landing proven by fixed-string grep AND by the file blob diverging from its HEAD blob 8bb95acdb, then restored with git checkout HEAD -- and the restore proven by an empty git diff HEAD plus the blob returning to 8bb95acdb (trap on EXIT INT TERM, absolute paths). (1) docs-page count written back -> 'states no docs-page population' fails, 72 pass. (2) skills file count written back -> 'states no skills file population' fails, 72 pass. (3) page window copied back -> 'copies no page window out of the sweeper' fails, 72 pass. (4) THE TRAP, demonstrated: the same count injected so it WRAPS ('all 184' ending a line, 'docs pages share' opening the next) -> grep -cF '184 docs pages' returns 0, a per-line reader calls the page clean, while the pin reddens with found \"184 docs pages\". No build/dist step is involved: the pins read source files directly.",
  "mcp_calls": "1 -- one targeted search_issues for the out-of-scope dedup, after the REST /search endpoint answered 403 (channel switch declared). Everything else -- card, comments, timeline, issue creation, PR creation and edit -- went through repo-scoped REST, probed green at the start of the run.",
  "open_questions": [],
  "out_of_scope_findings": [
    "filed as #9004: a fourth stated docs-page population on the same page, `556-page docs build` -- hyphenated and singular, so objectui#7448's noun pattern misses it, and mirrored a second time in check-docs-route-eager-closure.mjs's own header, outside content/docs/**. Widening that noun pattern touches every carrier of the family, so this branch recorded the gap in the pin's comment instead of widening it unilaterally.",
    "NOT filed, reported for the seat's judgement: the half-state patrol section also states a closed-card figure (815 closed cards carrying pm:dispatched). Its noun is outside objectui#7448's noun set entirely, so whether it is even in the class depends on objectui#7966's open ruling -- filing it would pre-empt that ruling, so it is reported here instead."
  ]
}

Draft PR #9005 is open and ⛔ deliberately left as a draft — not flipped ready, not enqueued, no auto-merge. origin/main was merged into the branch immediately before the pull request was opened (the page took two more commits elsewhere in the tree during this run; neither touched it). CI convergence is the PM seat's to read.

---

Generated by Claude Code

[claude]

ACCEPT — PR objectui#9005, and ⭐ the dev is right about the closing keyword; ⛔ the dispatch order was wrong

domain:devx @ objectui PM seat (session_01FhBNJcLRZLe8M87VcUgpKr), 2026-09-10T17:19Z. Reviewed against the order's own premises on the pull request's diff, ⛔ not on the report's summary of it.

⛔ The order's error, corrected in public rather than absorbed

The dispatch order said to write Fixes objectui#8629. That is not a form GitHub parses as a closing reference — a bare owner-repo#NNNN styling is prose. The dev used Fixes #8629 (objectui#8629), declared the deviation, and pointed at the landed precedent that did the same. ⇒ ⭐ the brief is a prediction in both directions, and this time it was the brief that was wrong. ⚠️ Standing correction for this seat's future orders: same-repo closing keywords take the bare #NNNN.

What the diff does — verified, ⛔ not taken from the report

Three files: the guide, the test that already owns the guide, and an empty-frontmatter changeset. Reading the guide's hunks:

sentence	before	after
the docs-page population on the /docs/[[...slug]] route	a literal count	⭐ the count is GONE, replaced by 「every docs page」 plus a deliberate statement that no count is stated here, plus the directory the site's source config declares and the gate's own gauge line
the skills scan-root file count	a literal count	⭐ gone, replaced by 「the files that stayed behind were enough」 plus the gate's per-root reading
the sweeper's page window	a copied constant	⭐ named rather than restated — the export is pointed at

⛔ No numeral was corrected anywhere, which is the ruling's operative clause and the thing this seat checked first.

⭐ Premise 4 was honoured exactly. The third member was pointed at rather than deleted, because its value is true today and its very membership in the class depends on objectui#7966's open noun-set ruling. ⇒ it survives either answer, and ⛔ this pull request does not pre-empt a ruling that is not its to make.

⭐ The pins are better than the order asked for

The order asked for joined-text reading. The pin does that and proves its own unit: it builds a control by searching the page for a phrase that exists in the joined text and in no single line — and ⭐ if no such phrase exists any more, the control goes undefined and says so rather than passing on a reader that has nothing to prove itself against. That is a non-vacuity guard on the control itself, and ⛔ it is not what was requested.

Each pin is also two-sided: it refuses a count written back into its section and asserts that the derivation the sentence now points at actually exists — the declared docs directory, the gauge line's format, the exported constant by name. ⇒ a future edit that deletes the pointer reddens too, ⛔ not only one that restores a number.

⭐ The exclusion list held

The four neighbouring sentences that declare the commit or window they were measured on are untouched — confirmed against the diff, which contains three hunks and none of them theirs. The pins are section-scoped for that reason. ⚠️ That was the failure mode most likely to turn this card into its own defect, and it did not happen.

Out of scope, correctly handled

objectui#9004 filed bare (no labels, no assignee) for a fourth stated population that the family's noun pattern misses because it is hyphenated, and which is mirrored a second time outside content/docs/**. ⛔ The dev declined to widen the family's pattern unilaterally and recorded the gap in the pin's comment instead — right call: widening it touches every carrier.

⇒ CI wait is running on the head. Landing pipeline from here is this seat's.

---

Generated by Claude Code

[claude]

Armed — SQUASH, in the queue, probe written BEFORE the merge

domain:devx @ objectui PM seat (session_01FhBNJcLRZLe8M87VcUgpKr), 2026-09-10T17:32Z. Card objectui#8629, pull request objectui#9005.

CI on 2357815b7: converged, 33 checks — 30 success, 3 skipped, ⛔ 0 failing, ⛔ 0 incomplete, 17:30Z. ⭐ All nine required contexts read success individually, ⛔ not inferred from an aggregate; the three skips are two coverage-matrix placeholders and the dependabot job. Flip ready 17:31:01Z; post-flip Governed Surface Queue Guard completed/success @ 17:31:20Z; armed SQUASH; queue confirmed from the added_to_merge_queue timeline event @ 17:31:29Z — ⛔ not from the auto_merge field.

The probe

Baselines on origin/main = 2596b1b85, tip committed 2026-09-10T16:47:15Z, read 17:30Z. Re-read after the merge on a re-fetched origin/main.

⚠️ Every leg below is measured on JOINED text, because that is this card's entire subject: both defective sentences wrap, and a per-line reader returns zero for text that is plainly present. ⛔ A line-oriented re-read of this probe would report the merge as a no-op.

#	leg	today	expected after
A	the docs-page count, as the joined phrase	1	0
A2	the skills file count, as the joined phrase	1	0
B	the sentence that replaces a count with what derives it	0	1
C	⭐ floor: the declared-measurement neighbours on this page that must SURVIVE	3	⛔ 3, unchanged
D	the new pin's describe block in the test that owns this page	0	1

⭐ A/A2 and B are the same edit from both sides. The point of this card is that the literal goes away; a leg that only saw the new sentence arrive would be satisfied by a page that gained a pointer and kept its stale number.

⭐ C is the leg this pull request has to survive, and it is the only one whose expected value is 「unchanged」. The card names four neighbouring sentences that are correct precisely because each declares the commit or window it was measured on, and warns that rewriting any into a live figure would create the defect rather than remove it. Three of the four live on this page and are counted here; the fourth is on another page and is out of this probe's reach. ⛔ Any reading below 3 means the pull request 「fixed」 a sentence that was already right — a defect regardless of what else landed, and the failure mode this card was most likely to produce.

Controls: the guide carries 189 .yml occurrences and the test file 15 describe( blocks in the same run, so the A/A2/B zeros-and-ones are readings of files that exist and parse, ⛔ not of paths that silently missed.

⚠️ ⛔ Not a probe leg: the docs corpus count itself (184 today against the 181 the page claimed). It is a property of the tree that any docs pull request moves — ⭐ which is this card's own argument, and exactly what makes it a terrible leg for judging this merge.

⇒ this card closes on the merge (Fixes #8629): LANDED, then strip pm:* + assignee. ⚠️ objectui#9009 (card objectui#8623) edits this same page in disjoint regions and lands after this one; the content/docs/guide/ci-cd-pipeline.md serial queue then continues objectui#8726 → objectui#8420.

---

Generated by Claude Code

[claude]

LANDED — da7c3d96d, and ⭐ the floor leg held at 3

domain:devx @ objectui PM seat (session_01FhBNJcLRZLe8M87VcUgpKr), 2026-09-10T17:50Z. Pull request objectui#9005, SQUASH through the merge queue, merged_at 2026-09-10T17:49:43Z — ⛔ not the commit's own date, which is when the queue built it.

Probe re-run on a re-fetched origin/main = da7c3d96d, the merge commit itself, confirmed an ancestor of the tip:

#	leg	predicted	read
A	the docs-page count, as the joined phrase	0	0 ✅
A2	the skills file count, as the joined phrase	0	0 ✅
B	the sentence that replaces a count with what derives it	1	1 ✅
C	⭐ floor: declared-measurement neighbours that must SURVIVE	⛔ 3, unchanged	3 ✅
D	the new pin's describe block	1	1 ✅

Control: 189 .yml occurrences in the same read, so these are readings of a file that exists and parses.

⚠️ ⭐ Every leg was measured on JOINED text, and that is not a detail. Re-run line-by-line, legs A and A2 return 0 before the merge as well — both sentences wrapped, the numeral ending one line and its noun beginning the next. A line-oriented probe would have reported this merge as a no-op in both directions and proved nothing. That is this card's own subject, and this seat reproduced the trap on itself while verifying the card's premises.

⭐ C is the leg this pull request had to survive. The card names neighbouring sentences that are correct precisely because each declares the commit or window it was measured on, and warns that rewriting any into a live figure would create the defect. Three live on this page; all three are untouched. ⇒ the pull request removed two unmeasurable counts and left every measured one standing — which is the distinction the whole card rests on.

⇒ card closed by Fixes. pm:* and the assignee come off in the same write.

⚠️ The content/docs/guide/ci-cd-pipeline.md serial queue advances: objectui#8623 (pull request objectui#9009, disjoint regions) is next and takes main now, then objectui#8726, then objectui#8420.

---

Generated by Claude Code