Found while fixing objectui#7886 (content/docs/guide/ci-cd-pipeline.md published the same two hand-copied document counts objectui#7825 had just removed from check-links.yml's header). Measured, not inferred. Reported rather than built, per that card's binding scope.
What
objectui#7448's document-count pin now exists in three copies, and all three read .github/workflows/*.yml headers:
| pin |
file it reads |
scripts/__tests__/check-doc-fence-languages.test.ts |
.github/workflows/doc-fence-languages.yml header comments |
scripts/__tests__/check-doc-component-types.test.ts |
.github/workflows/doc-component-types.yml header comments |
scripts/__tests__/check-links-workflow.test.ts (objectui#7825) |
check-links.yml header comments and its own leading block comment |
Nothing reads content/docs/**. The class's worst-placed instance so far was therefore invisible to every copy of its own pin: prose the site publishes to readers, not a comment contributors read.
The measurement
The shape is exactly the class. Both regexes fire on the pre-fix paragraph of content/docs/guide/ci-cd-pipeline.md and return nothing on the rewritten one:
PRE-FIX ci-cd-pipeline.md "Link Checking" paragraph:
twin regex (objectui#7448) => ["183 pages","15 files"]
objectui#7825 regex => ["183 pages","15 files"]
POST-FIX (objectui#7886, PR #7899):
twin regex => []
objectui#7825 regex => []
And nothing was watching. Nine test files read content/docs/guide/ci-cd-pipeline.md — ci-cd-pipeline-doc, check-doc-expression-carriage, check-doc-example-shared-reader, check-handler-key-read-sites, check-skill-examples, check-skill-eval-tokens, merge-queue-reporting, lint-workflow, doc-version-claims — and not one asserts anything about either figure or the prose carrying them. A repo-wide grep for the two literals found them only on that page and, as deliberate historical fixtures, inside check-links-workflow.test.ts.
Why it is worth recording rather than fixing on the spot
A fourth copy is not another test file beside a gate: it introduces a new scan population over the content/docs tree, which this lane requires to land report-only first, with the census as the deliverable. That is a card of its own, not a rider on a two-paragraph prose PR — the same reasoning objectui#7825 used when it declined to take the page as a rider.
Not independent of objectui#7888
objectui#7888 is the narrowness problem in the same family: the twins' regex requires the numeral and the noun to be adjacent, so 15 INTERNAL documents scores clean. This finding is the coverage problem: which surfaces the family reads at all. They are separable, and the order matters — widening a pattern that reads no content/docs file changes nothing here, and pointing a pattern at content/docs while it still misses one adjective would land a census that under-reports on its first run. Whoever picks either up should read the other first.
Blocked-by: nothing. It can be worked before or after objectui#7888; doing objectui#7888 first makes the census this one needs honest on its first run.
Back-links
objectui#7886 (this page's fix, PR #7899), objectui#7825, objectui#7448, objectui#7888.
Generated by Claude Code
Found while fixing objectui#7886 (
content/docs/guide/ci-cd-pipeline.mdpublished the same two hand-copied document counts objectui#7825 had just removed fromcheck-links.yml's header). Measured, not inferred. Reported rather than built, per that card's binding scope.What
objectui#7448's document-count pin now exists in three copies, and all three read
.github/workflows/*.ymlheaders:scripts/__tests__/check-doc-fence-languages.test.ts.github/workflows/doc-fence-languages.ymlheader commentsscripts/__tests__/check-doc-component-types.test.ts.github/workflows/doc-component-types.ymlheader commentsscripts/__tests__/check-links-workflow.test.ts(objectui#7825)check-links.ymlheader comments and its own leading block commentNothing reads
content/docs/**. The class's worst-placed instance so far was therefore invisible to every copy of its own pin: prose the site publishes to readers, not a comment contributors read.The measurement
The shape is exactly the class. Both regexes fire on the pre-fix paragraph of
content/docs/guide/ci-cd-pipeline.mdand return nothing on the rewritten one:And nothing was watching. Nine test files read
content/docs/guide/ci-cd-pipeline.md—ci-cd-pipeline-doc,check-doc-expression-carriage,check-doc-example-shared-reader,check-handler-key-read-sites,check-skill-examples,check-skill-eval-tokens,merge-queue-reporting,lint-workflow,doc-version-claims— and not one asserts anything about either figure or the prose carrying them. A repo-wide grep for the two literals found them only on that page and, as deliberate historical fixtures, insidecheck-links-workflow.test.ts.Why it is worth recording rather than fixing on the spot
A fourth copy is not another test file beside a gate: it introduces a new scan population over the
content/docstree, which this lane requires to land report-only first, with the census as the deliverable. That is a card of its own, not a rider on a two-paragraph prose PR — the same reasoning objectui#7825 used when it declined to take the page as a rider.Not independent of objectui#7888
objectui#7888 is the narrowness problem in the same family: the twins' regex requires the numeral and the noun to be adjacent, so
15 INTERNAL documentsscores clean. This finding is the coverage problem: which surfaces the family reads at all. They are separable, and the order matters — widening a pattern that reads nocontent/docsfile changes nothing here, and pointing a pattern atcontent/docswhile it still misses one adjective would land a census that under-reports on its first run. Whoever picks either up should read the other first.Blocked-by:nothing. It can be worked before or after objectui#7888; doing objectui#7888 first makes the census this one needs honest on its first run.Back-links
objectui#7886 (this page's fix, PR #7899), objectui#7825, objectui#7448, objectui#7888.
Generated by Claude Code