fix(docs-audit): give the route:/client: lead one character class

[os-steve]

Fixes #11494

scripts/docs-audit/affected-docs.mjs had eight scans that ask "is a route: / client: declaration written here?". Seven spelled the run between the colon and the value \s*; declarationsIn spelled it [ \t]*. One character class, one character of difference — a newline — and a declaration whose value sits on the next line was seen by both, in two different buckets, so one declaration reached the denominator twice.

Re-derived on this base, not quoted from the card

The card's numbers are from 951b025e. #11500 (2d1290ae) has since rewritten declinedIn in this exact file, so "byte-identical on origin/main" is a statement about a tree that no longer exists. Re-derived on cd932772:

	before	after
A card fixture — double-quoted value on the next line	rows 1 · routesDeclared 3 · declined 2 · brokenScan 1	rows 1 · routesDeclared 2 · declined 1 · brokenScan 1
B single-quoted value on the next line	rows 1 · routesDeclared 2 · declined 1 · brokenScan 1	rows 1 · routesDeclared 1 · declined 0 · brokenScan 0
C wrapped client:	clientsDeclared 2 · declined 1 · brokenScan 1	clientsDeclared 1 · declined 0 · brokenScan 0
D wrapped non-literal	declined 1, named route:	declined 1, named route: ROUTES.health

A reproduces the card exactly, second entry and all: { key: 'route', line: 2, text: 'route: ' } — an empty value, naming nothing a reader can act on.

B is the louder half the card did not measure, and it is what settles which class wins. The row recognizer already spells \s*, so a wrapped single-quoted value is read as a row — and was then billed unread by the same file, raising a PARTIAL-read verdict with exit 1 on a wholly accurate ledger. That is the false-red direction, which this file prices as costing the same trust a false green does. Making [ \t]* win instead would have had to move the recognizer too — a change to the measured population, which the header of --bridge-coverage attaches a before/after standard to — and would have kept the empty-valued entry as the surviving one.

Note that rows + declined === routesDeclared kept balancing throughout (1 + 2 === 3). The partition is blind to this by construction: what moved was the population, not the arithmetic.

Route taken: (2), not (3)

Route 3 (claim once by index) is the larger change and strictly weaker here. It suppresses the symptom at each site — the double bill in A is declinedIn + unreadableIn, but in B and C it is the recognizer + unreadableIn, so three separate claim registrations would be needed — and it leaves the two scans still disagreeing about the character class. Route 2 removes the disagreement at the root and fixes all four shapes with one line.

Written once

declLead is now the only place the class is spelled, and all eight lead scans are built from it. The other seven are byte-identical in source and flags — verified mechanically, 1 of 8 changed:

CHANGED    declarationsIn   built=/\b(route|client)\s*:\s*/g   orig=/\b(route|client)\s*:[ \t]*/g
IDENTICAL  declinedIn       /(route|client)\s*:\s*(["`])([^\n]{0,120})/g
IDENTICAL  routeRe · nextRouteRe · windowClientRe · routesDeclared · codeLeads · outsideCode

Priced free on today's tree

Zero live instances re-checked here, not inherited: seven ledgers (unchanged), and grep -nE '\b(route|client)[ \t]*:[ \t]*$' over all seven returns 0 hits (grep exit 1). --bridge-coverage delta across the change:

rowsParsed 268→268 · routesDeclared 268→268 · clientsDeclared 222→222
reachable 45→45 · clientRows 222→222 · unreachable 177→177 · leadsOutsideCode 0→0
brokenScan 0→0 · unreachableRows byte-identical: true

Non-vacuity — both directions, mutation proven on disk

--self-test: 366 → 381 cases. Two ablation legs, each with the mutation confirmed on disk by grepping for the removed and injected text (not an editor exit code), and each restored byte-identically by git hash-object:

• Leg A — the faithful pre-fix state (declarationsIn reverted to its inline [ \t]*): 8 checks fail, reproducing the card's numbers verbatim — got "1 row / 3 route / 2 declined" and got "line 2: route: \"GET /api/v1/gone\", line 2: route: ". The partition check stays green throughout, which is the point.
• Leg B — the shared class reverted in declLead (all eight scans at once): 8 checks fail, including row → got "undefined → undefined" — which proves every one of the eight call sites really does read the helper rather than keeping an inline copy.
• Restore leg for both: hash-object byte-identical to pristine, --self-test exit 0, 381 pass.

Boundary deliberately not crossed

declLead unifies what may sit between the colon and the value. The key part is still each call site's own: declarationsIn anchors with \b, the other seven do not, so subroute: 'GET /x' is a declaration to seven of eight and mints a silent phantom row (rows 2 on a file declaring one, brokenScan 0). Unifying that moves the measured population, so it is filed as #11542 and pinned here — the card that closes it moves the pin instead of finding none.

Gates

Derived with node scripts/pm/dispatch-gates.mjs --repo objectstack-ai/objectstack (the script derived the change set itself from the merge base; provenance line asserted objectstack). Run at e8a9b04e:

gate	verdict line
check:docs-audit-scope	✓ affected-docs self-test: 381 cases pass. / ✓ docs-accuracy-audit scope is in sync
check:cross-package-test-inputs	OK: 14 package(s) read outside themselves, all declared
check:entry-guard	✓ check:entry-guard: 139 scripts/ file(s)
check:parse-guard	pass
check:pm-governed-merges	✓ check-governed-merges --self-test: 129 assertions
check:pnpm-filter-targets	✓ check-pnpm-filter-targets --self-test: 40 assertions
check-ci-filter-parity.mjs	OK: all 89 declared cross-package glob(s) (77 unique) are covered
check-affected-docs.mjs	pass
check-nul-bytes.mjs	check-nul-bytes: OK (scanned 6460 text file(s) … no raw ASCII control bytes)

Repo-wide pnpm lint (eslint . --no-inline-config) was run in full, not narrowed: os-verify-lock: VERDICT command-exit 0 · held the lock 65s. pnpm exec eslint --format json on the changed file alone reports 1 file, 0 errors, 0 warnings.

No changeset: this edits a CI-internal script and releases nothing — the textbook skip-changeset case, in lint.yml's own words. Label applied additively and read back.

---

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.