docs(docs-audit): record that the periodic full-audit backstop is not running

[os-steve]

Fixes #9231

Answer: the backstop does not exist

This card was a verification first. The answer is no — there is no live schedule running the docs-accuracy --all full audit, and there never was one.

scripts/docs-audit/README.md §4 described a cron routine on a monthly / per-release cadence. Measured 2026-08-18, four independent ways, all negative:

check	result
GitHub Actions workflows	No workflow runs this audit at all. 31 workflows registered on the repo; the 10 carrying a schedule: trigger are all unrelated.
Routines (list_triggers)	The only cron Routine on the agent-seat account is the hourly triage seat, 18 * * * *. No docs-audit Routine, enabled or disabled.
The named creation mechanism	The schedule skill §4 pointed at is not in .claude/skills/.
Trace of any periodic run	Exactly three docs-accuracy audit PRs in repo history (#3243, #4219, #4312), each hand-initiated against a named issue family, most recent 2026-07-31. No cadence.

Two notes on method, because the card was explicitly about not reading a conclusion off a field that cannot carry it:

• The scheduled-workflow list was derived, not copied — and copying it was wrong. The first push of this PR said 11 and included cut-rc, inherited from an earlier triage comment on the card. On main, cut-rc.yml has no schedule: trigger; its only schedule: match is a comment saying it deliberately has none, which an unanchored grep counts. Ten workflows carry a real one. The conclusion is unchanged, but consuming a stated list as authoritative is precisely the failure this card measures, so the README now carries the derivation command rather than the list alone.
• The workflow list was read from the Actions API, not from disk. The two disagree in both directions: matrix-aggregate-experiment.yml is registered in Actions with no file on main today, and a scheduled workflow auto-disabled after 60 days of inactivity leaves its file byte-identical. Neither view alone answers "is it running".
• The Routines surface is readable by an agent. The earlier triage note on this card recorded that Routine state is "readable only from the maintainer's Routines UI — no agent-facing surface can confirm it". That is falsified: list_triggers enumerates every Routine the account owns, and a cron Routine is never hidden by the include_completed filter, so its absence is a real reading rather than a listing artifact.

Second finding: change-scoped is not a backstop

Independent of whether it ran. §4 scoped the routine to "the change-scoped doc list since the last audit", while the cost note in the same file calls part 4 the "periodic full backstop". Those are different runs and only the second is a backstop — a change-scoped list is derived by the very anchor heuristic whose misses the backstop exists to catch, so scoping it that way re-inherits the blind spot it is meant to cover.

So the leg was doubly unsound: absent, and mis-specified even as written. For the record, --all itself is sound as a coverage mechanism — it enumerates all 178 hand-written docs, verified against the audit workflow's own inline scope via check-audit-scope.mjs.

Third finding: the named fallback is not a backstop either

The card names a cheap fallback — point the audit's args wiring at packageMentionDocs. It was not implemented (see below), and it also would not work. Measured across the 8 most recent packages/**-touching commits on main, the coarse set is wider but not a superset:

commit	precise docs	packageMentionDocs	in docs but NOT in packageMentionDocs
a4331227b	5	21	3
48032c927	9	31	3
19539b4b2	41	119	6
40d5b2d4c	45	120	6
42d899071	47	120	7

In every one of the 8, between 3 and 7 pages the precise scope correctly finds are absent from the coarse set — api/error-catalog.mdx, automation/approvals.mdx, api/plugin-endpoints.mdx, data-modeling/relationships.mdx among them. That is #9192's own finding restated: the two sets miss in different directions, because the coarse one is a dependency-graph proxy and pages documenting a change through the SDK surface never name the implementing package.

So the fallback trades one incomplete set for another and loses pages the current scope gets right. Only --all is a backstop. This is a fact the PM should have before revisiting the ruling, not a recommendation.

What this changes

Documentation only. §4 now states the measured fact, carries the four checks so the next person re-reads them instead of re-deriving them, and keeps the intended design for whoever stands the routine up. The cost note no longer names a backstop that is not running.

⛔ Deliberately not done, both being the maintainer's / PM's call:

• No scheduled full-audit workflow or Routine was created or re-enabled — a periodic LLM audit spends real budget on a cadence.
• The cheap fallback the card names (pointing the audit's args wiring at packageMentionDocs) is not implemented. Choosing it means revisiting the The docs-drift advisory lists pages by package dependency, so it is wrong in BOTH directions — measured: 2 of 3 listed pages irrelevant, and the 2 pages that actually document the changed surface were not listed #9192 option-A ruling, whose second leg this PR shows is absent.

scripts/docs-audit/affected-docs.mjs was not touched — it is held by #9294.

Verification

Gate union re-run on the final commit 673885abd, working tree clean:

node scripts/docs-audit/check-affected-docs.mjs
  ✓ affected-docs self-test: 197 cases pass.
node scripts/docs-audit/check-audit-scope.mjs
  ✓ docs-accuracy-audit scope is in sync with content/docs/: 178 hand-written doc(s).
  ✓ release-owned pages are in scope and read-only: 9 page(s) under content/docs/releases/
node scripts/check-nul-bytes.mjs
  check-nul-bytes: OK (scanned 6122 text file(s) ... no raw ASCII control bytes)

Gate family derived from the changed path with node scripts/pm/dispatch-gates.mjs scripts/docs-audit/README.md, which named check-affected-docs.mjs via the scripts/docs-audit/** trigger in docs-drift-check.yml.

No changeset: this is an internal tooling README, publishing nothing.

---

Generated by Claude Code

---

Generated by Claude Code

[github-actions]

📓 Docs Drift Check

Nothing in this diff resolved to a documentable surface (no symbol, route or SDK anchor derived from 0 changed package(s)), so this run has no opinion about the docs.