[finding] docs-audit: the ledger row recognizer now reads `codeOnly` but still not `typeDeclRegions` — a literal-union `route:` TYPE member still mints a silent phantom row

[claude]

Filed unassigned by the os-dev seat implementing #10683. Recording, not fixing — measured as a by-product of that card's before/after, and it is a second discriminator with its own population move, which the #10683 card's own rule says wants its own before/after rather than a rider. No live instance on today's tree.

What

#10683 moved parseLedgerSource's row recognizer (and the first routesDeclared term) to read through codeOnly, closing the comment/string-payload phantom. But #10500 built two exact discriminators for "is this a declaration in code position", and declarationsIn reads through both:

const code = codeOnly(text);
const skip = typeDeclRegions(code);

The recognizer now reads the first and still not the second. codeOnly does not blank type declarations — it has no reason to, they are code — so a single-quoted route: inside an interface / type body is still a row.

Measured (this branch, parseLedgerSource driven directly)

export interface Entry { route: 'GET /api/v1/gone' | 'GET /api/v1/meta'; client: string }
export const L = [
 { route: 'GET /api/v1/meta', family: 'metadata', disposition: 'sdk', client: 'meta.getTypes' },
];

⇒ rows 2 · routesDeclared 2 · clientsDeclared 1 · declined 0 · outsideCode 0

Two rows from a file declaring one — the identical silent shape #10683 fixed, arriving through the other discriminator. outsideCode (#10683's reporting half) does not see it either: the lead is in code position, so it is correctly not a prose lead. And no verdict fires, for the same reason as before — rows and routesDeclared move together, so the partial-read guard sees no gap.

The literal-union member is not hypothetical as a shape: orphanSource in the self-test pins exactly that spelling on the client: key (client: 'i18n.getLocales' | 'i18n.getPlurals'), written there specifically because "a literal-union TYPE member opens with the very quote this counter reads". The route: key is the same spelling with nothing keeping it out of the row loop.

Why it is a finding and not a live defect

All seven live ledgers declare route: string; and nothing else — scanned on this branch: 0 single-quoted route: leads inside a type declaration, across all seven. Live figures stay exact (259 of 259, 221 of 221, 176 unreachable).

Direction, not a prescription

Applying typeDeclRegions to the recognizer is one line beside the codeOnly call and reuses the existing region list. ⚠️ It moves the measured population on the same terms #10683 did, so it wants its own before/after against the --bridge-coverage header — which is exactly why it was not ridden along on that card. Measured delta today would be zero.

The alternative reading is that the row loop should stop deciding code position for itself and filter declarationsIn(text) — the file already states that design goal twice ("two scans that each decide separately what "in code position" means are two scans that can drift into disagreeing while both look right"). That is a bigger change with the same population effect.

Refs: #10683, #10500, #9896.

---

Generated by Claude Code

[huangyiirene]

Triage first-touch: promoted to queue, type Task. This is the silent direction of the pair (#10794 is the loud one and was closed not-planned): a literal-union route: type member mints a phantom row with no verdict — the identical shape #10683 fixed, one discriminator over. Scope: apply typeDeclRegions beside the codeOnly call in the row recognizer (one line, reuses the existing region list), with its own before/after against the --bridge-coverage header — expected delta today is zero, and that measurement is part of the deliverable, not optional. The wider refactor (row loop filtering declarationsIn(text)) is out of scope unless the dev shows the one-liner can't hold. Size/model suggestion: S, sonnet.

---

Generated by Claude Code

[claude]

PM claim — dispatching now.

• session: c970724d-303c-5614-9d20-a3f92205cfad
• branch: claude/issue-10793-typedecl-regions-recognizer
• seat: domain:devx

Verified before dispatch, on origin/main @ 7c02a4529c:

• The surface is scripts/docs-audit/affected-docs.mjs (3389 lines). declarationsIn at :1328-1330 reads through both discriminators; parseLedgerSource at :1482 reads through only codeOnly at :1518.
• The gap is already self-documented in the file — :1512 carries ⛔ ONE of #10500's two discriminators, not both. typeDeclRegions is still NOT applied. So the card's premise is confirmed by the code's own comment, not only by the card.
• No collision: none of the five open/queued PRs (docs(qa-checklist): correct two knownGaps that still deny the provisioning.use resolve #10883 fix(scripts): the required-context self-test proves WIRING with invocations, not prose (#10877) #10884 fix(ci): collect every self-test verdict in a step, instead of masking after the first failure #10887 fix(devx): stop check:published-files spelling a bare repo root it never opens #10890 fix(devx): assert that a published README's repo-relative targets exist #10892) touches affected-docs.mjs.

The card is explicit that there is no live instance — all seven ledgers declare route: string; — so this is a latent-shape fix, and the agent is told to hold the population-move discipline the card asks for rather than to hurry.

If an earlier claim comment with a different session ID appears above this one, that claim wins and this one is void.

---

Generated by Claude Code