check_workflow_job_banners.py's docstring still names suppression-guard.yml, which #635 deleted

[Yaraslaut]

scripts/check_workflow_job_banners.py's docstring, in its "Scope" paragraph,
still lists suppression-guard.yml among the single-job workflows that have
never used the # ── banner style:

scripts/check_workflow_job_banners.py:50:suppression-guard.yml and the two wasm workflows are single-job files that have

That workflow no longer exists. #635 deleted it when it folded the
NOLINT-directive gate into drift-guard.yml's prose-lint job.

Verification status: reproduced on master @ c55ea5b7

The file is gone:

$ ls .github/workflows/
ci.yml
docs.yml
drift-guard.yml
mutation.yml
spec-sync.yml
wasm-demo.yml
wasm-ladder.yml

And the docstring is the only surviving reference to it anywhere in the tree:

$ grep -rn "suppression-guard" --include='*.py' --include='*.sh' --include='*.yml' --include='*.md' . | grep -v '^./.git/'
scripts/check_workflow_job_banners.py:50:suppression-guard.yml and the two wasm workflows are single-job files that have

Not verified: nothing about the gate's behaviour. The list in that
paragraph is prose explaining a rule that is derived from each file at run time
("Workflows that contain no banner at all are skipped, and that is derived from
the file rather than listed here"), so a name in it is documentation, not data,
and python3 scripts/check_workflow_job_banners.py . passes today. I did not
read the checker looking for any other use of the name; grep says there is
none.

Why it is worth a line

The defect is small and the irony is the point: a stale fact inside a gate
whose whole subject is stale facts. It sits in the paragraph a reader goes to
when asking "why is my new single-job workflow not being checked?", where a
name that no longer resolves makes the derivation rule harder to trust rather
than easier.

Found while reading the two most recent lint patterns before writing #628's
raw-bidi-control gate. Filed rather than folded, per AGENTS.md: it does not
belong to that change.

What would change the verdict

Close when the name is removed from the paragraph, or replaced by one of the
files that is actually in the tree. Re-open if suppression-guard.yml comes
back.

🤖 Generated with Claude Code

https://claude.ai/code/session_01VptDWG2fKr2vBnLSJcgzgW

[Yaraslaut]

Triage: valid — reproduced on f7c231df, and it is the only one left

$ git grep -n "suppression-guard" -- .
scripts/check_workflow_job_banners.py:50:suppression-guard.yml and the two wasm workflows are single-job files that have

$ ls .github/workflows/suppression-guard.yml
no — deleted by #635

One surviving reference, in a docstring, to a workflow that no longer exists.

Gate behaviour is unaffected, and that is worth stating rather than leaving implied. The docstring is explaining why certain files are skipped, but the skip is derived per file — "workflows that contain no banner at all are skipped, and that is derived from the file rather than listed here" — so the name is illustrative, not load-bearing. I verified the derivation holds during the #635/#638 landing sweep: the checker run against a tree with suppression-guard.yml already deleted exited 0 and simply stopped printing a line for it.

So this is a documentation defect with a real cost and no functional one: the example that explains the rule cites a file the reader cannot open, which makes the rule harder to check rather than easier.

triage: valid, documentation, area: ci.

For whoever takes it: the docstring lists five files by name as examples of the single-job no-banner style. Replacing suppression-guard.yml with another current example keeps the sentence useful; deleting the name and leaving "and the two wasm workflows" leaves the count wrong. Either way the list is exactly the kind of prose that rots — worth a line saying it is illustrative, so the next deletion does not produce a third round of this.

Related: #645 (nothing gates a prose job-id reference inside a workflow comment) is the same class one level over — a stale name in a comment that no gate can see. Both are cheap; neither is urgent.

🤖 Generated with Claude Code

https://claude.ai/code/session_01VptDWG2fKr2vBnLSJcgzgW