fix(lint): one-line fix lines for 17 more author-time rule ids (slice 16); os explain RULE_ID carries what they explained

[objectstack-fleet]

Part of #22161
Clause-②: no

Stage 2 of the card, slice 16: the over-long fix: lines, continued in slices 11 to 15's shape. Each finding's hint (the CLI's fix: line, the runtime issue's hint) of 17 more packages/lint rule ids, in twelve rule files, is now one remedy line of at most 200 characters, and what the lines used to explain is each id's RULE_EXPLANATIONS entry behind os explain RULE_ID. The card stays open: 24 author-time ids still carry a hint over 200 characters (the last section).

What changes

• One remedy line per finding. The 17 ids, every hint the lint suite fires, before → after (count of distinct hints · length range):

rule id	file	lint suite, before → after	runtime door
approval-approver-not-membership-tier	validate-approval-approvers.ts	2 · 297–313 → 3 · 130–168	flow write, 2xx advisory
approval-expression-no-empty-policy	same	1 · 294 → 1 · 121	flow write, 2xx advisory
approval-expression-invalid	same	5 · 102–292 → 5 · 90–144	flow write, 422 (the resolveAs arm a 2xx advisory)
dataset-include-unknown	validate-dataset-references.ts	4 · 188–336 → 2 · 94–97	dataset write, 422
dataset-field-unknown	same	9 · 164–206 → 6 · 131–139	dataset write, 422
title-format-retired	validate-record-title.ts	1 · 267 → 1 · 148	CLI only
title-unresolvable	same	1 · 237 → 1 · 122	CLI only
list-view-field-unknown	validate-list-view-field-refs.ts	4 · 94–258 → 4 · 57–62	view / object write, 422
list-view-field-dotted	same	4 · 145–233 → 4 · 97–158	view / object write, 422
translation-target-unknown	validate-translation-references.ts	85 · 45–317 → 30 · 45–164	CLI only
seed-insert-mode-duplicates-on-replay	validate-seed-replay-safety.ts	1 · 309 → 1 · 131	CLI only
chart-measure-unknown	validate-chart-bindings.ts	9 · 91–285 → 8 · 78–115	report write, 422 (presentation positions a 2xx advisory)
object-field-ref-unknown	validate-object-field-refs.ts	25 · 155–284 → 8 · 116–186	object write, 422
view-container-shape	validate-view-containers.ts	3 · 158–256 → 3 · 158–187	CLI only
seed-value-outside-state-machine	validate-seed-state-machine.ts	4 · 247–251 → 5 · 135–139	CLI only
ai-skill-surface-mismatch	validate-ai-surface-affinity.ts	6 · 218–235 → 7 · 126–171	CLI only
page-source-className-tailwind	validate-page-source-styling.ts	2 · 195–231 → 2 · 127–167	CLI only

A count that falls is a list that no longer makes each hint distinct; a count that rises is a case this slice added (a long position, object, skill or page name fired to read the bound on it).

• Lists are no longer echoed (slices 11, 12 and 14's rule: a list the verdict's "Did you mean" already covers).
  • dataset-include-unknown, dataset-field-unknown, list-view-field-unknown and object-field-ref-unknown no longer append the object's whole field list (account.detail); chart-measure-unknown no longer lists the dataset's measures, and keeps its own "Did you mean" after the remedy.
  • translation-target-unknown no longer lists the declared names on any leg. Two facts a list used to carry stay as one bounded clause: a stale _views key is told that a nameless default list is keyed default (the packages/lint:collectViewRecord 收窄 _views 键到运行时裸键单拼写 —— #5164 裁 A 的 lint 段 #6038 steer the declared-view list gave a _views.list key), and a _tabs key over an object with no preset bar is told to declare the tab in a page's interfaceConfig.userFilters.tabs, because a list view's own tabs has no renderer (the mis-fix the old tail guarded).
• The reasoning moves, the remedy stays.
  • The approval ids: what the three expression roots hold, why record and previous are unbound at an approval node, and what resolveAs does are the new approval-expression-invalid entry; what each onEmptyApprovers policy does is the new approval-expression-no-empty-policy entry (the nudge still names all four policies); the membership-tier line names the value once and points at the tiers the verdict lists, and why a business role is always a position was already the entry's.
  • The dataset ids: the three-hop limit, which field types join, that a.b includes a, and that a join is derived from the object graph (no ON clause) are the entries'.
  • The title ids gain entries: why titleFormat is retired (render-only, neither returned nor queryable, nameField wins) and how a record title resolves, with its Record #ID fallback.
  • list-view-field-dotted keeps both doors' remedies in their own words ("denormalise the value onto … and name that"); what a denormalised field is moves to the entry, in a paragraph naming each door's ways out.
  • chart-measure-unknown: what series[].name and yAxis[] carry at a presentation position, and that none of these surfaces has a suppressWarnings key, are the entry's.
  • The seed ids: the externalId examples and that the seed exemption never licenses an undeclared state are the entries'.
  • view-container-shape: that machine-assembled manifests carry a ViewItem under viewItems: was already the entry's.
  • ai-skill-surface-mismatch: when surface: 'both' is right is the entry's.
  • page-source-className-tailwind gains an entry: why a Tailwind class in page source produces no CSS, the theme-token examples and the drawer/modal overlay advice.
• Five ids gain an os explain entry, so their rule: line now carries the pointer: approval-expression-invalid, approval-expression-no-empty-policy, title-format-retired, title-unresolvable and page-source-className-tailwind. Each already had a rule id constant on the root barrel, so no published name is added (the closure reading below). Ten existing entries gain text. Read from the built tables, none removed: 146 → 151 against the base d206d08ca, 145 → 150 against main at e540e7ffb (which feat(spec,service-automation,lint)!: a CRUD node's filter operands are value slots, the query placeholders stay — the {…} token refused (#19939 pass 4, S6) #23154 left one entry shorter).
• Enumerations held to their sources. An entry that spells a list out is pinned to it by the rule's test: the empty-slate policies (and the nudge's) to ApprovalNodeConfigSchema's onEmptyApprovers enum, the expression roots to APPROVER_EXPRESSION_ROOTS, and the title-eligible types to TITLE_ELIGIBLE_TYPES.
• Nothing else moves.
  • Every message, rule id, severity and path is unchanged, and so is what each rule accepts or refuses.
  • Every hunk in the twelve rule files is a hint expression, a comment, or the removal of the now-unused SHAPE_HINT table in validate-chart-bindings.ts. Two hint expressions choose between two remedy texts on a fact the rule already read: the _tabs leg on whether the object has a preset bar (it chose between two list tails before), and checkFieldPath keeps the field list for dataset-filter-field-unknown, outside this slice.
  • .changeset/22161-lint-slice-16-one-line-fix.md: @objectstack/lint patch, Clause-②: no, with the runtime hint FROM → TO for every id that fires at the runtime gate.

File surface: the claim's

The claim (6137662991) names the twelve rule files, rule-explanations.ts, the shared one-line producers where an id converges on them, a text-only test rider in packages/cli / packages/metadata-protocol, and the changeset.

• The edits stay inside it: the twelve rule files, their twelve test files, rule-explanations.ts and the changeset.
• system-fields.ts and checked-parse.ts needed no edit: none of the 17 is an anchor or a parse id.
• No packages/cli or packages/metadata-protocol rider was needed: no test there reads these hints' text, and both suites pass unchanged on the branch (below).

The runtime wire

Measured at the door with runRuntimeAuthoringRules, on the base dist and on the rebuilt one: a scratch probe of 20 writes across flow, dataset, object, view, report, seed, page and agent.

• a flow write with an approval node:
  • approval-approver-not-membership-tier (2xx advisories[].hint): 313 → 138;
  • approval-expression-invalid (422 issues[].hint): an empty expression 216 → 90, a record root 292 → 144; a parse failure (129), a bare root (110) and the resolveAs advisory (102) unchanged;
  • approval-expression-no-empty-policy (2xx advisory): 294 → 121.
• a dataset write: dataset-include-unknown (422), an entry naming nothing / an ordinary field, 239 / 187 → 93 / 93; dataset-field-unknown (422), a dimension / a measure, 184 / 182 → 132 / 130.
• an object write: object-field-ref-unknown (422) 180 → 128 — the same text is the fix of the publish receipt's object_field_ref_unknown runtime probe issue (build-probes.ts); on its listViews, list-view-field-unknown (422) 110 → 57 and list-view-field-dotted (422) 231 → 157.
• a view write: list-view-field-unknown (422) 110 → 57; list-view-field-dotted (422), a projection / a filter, 231 / 143 → 157 / 95.
• a report write: chart-measure-unknown, 422 at the query position 117 → 106, 2xx advisory at a presentation position 301 → 113.
• No write reaches the other eight: the two title ids, both seed ids, view-container-shape and page-source-className-tailwind are CLI_ONLY in the registry, and translation-target-unknown and ai-skill-surface-mismatch are reference-suite members on the frozen flow default (a flow snapshot carries neither translations nor skills).

Readers of hint, read-only against objectui main at fd4872f9 (fetched into a ref of this session's own, git grep for .hint reads):

• Studio's saveAdvisoryToast.ts formatFinding renders the rule, where, message and hint for every 2xx advisory, so the three approval advisories and the chart presentation advisory shorten there.
• DraftChangesPanel.tsx reads hint only for its in-browser validateSecurityPosture pass (none of these ids).
• No objectui source reads a 422 issue's hint.
• REST (PUT /api/v1/meta) forwards issues / advisories as JSON (toIssue copies hint verbatim). packages/mcp registers no metadata write tool (its tools are the record, action, run and expression tools), so an MCP or AI author reaches the gate through REST.
• No reader in either repository matches on hint text.

Census (taken first, before any edit, at the base d206d08ca)

Method: the earlier slices' scratch preload (NODE_OPTIONS=--import, never committed), which records every finding-shaped object, deduped by (rule, hint) per process, with its push site. Lengths are the hint alone.

• packages/lint suite: the 17 ids re-measured at the base equal the claim's lengths exactly (313, 294, 292, 336, 206, 267, 237, 258, 233, 317, 309, 285, 284, 256, 251, 235, 231).
  • Base: 240 ids fire, 65 over 200 = 21 fenced + 3 repo-gate + 41 author-time. The 41 are slice 15's remaining list, id for id.
  • After, at 0a66efe55 (before the merges): 240 ids fire, 48 over 200 = 21 fenced + 3 repo-gate + 24 author-time, and every hint of the 17 is at most 200 (the table). The 24 are the 41 less these 17.
  • At the final head abe4ce231 (after both merges, rebuilt): 239 ids fire (feat(spec,service-automation,lint)!: a CRUD node's filter operands are value slots, the query placeholders stay — the {…} token refused (#19939 pass 4, S6) #23154 retired one), 48 over 200, the same 21 + 3 + 24, and the 17 ids' lengths equal the pre-merge census id for id. main at e540e7ffb, built in the base worktree, fires the 17 at exactly the base lengths.
• Runtime publish gate, the whole @objectstack/metadata-protocol suite (which reads @objectstack/lint from source), in a base worktree and at the branch: approval-expression-no-empty-policy 294 → 121, dataset-field-unknown 188 → 136, list-view-field-unknown 102 → 58, object-field-ref-unknown 171–202 → 128–161; approval-expression-invalid fires there only on its parse arm (129, unchanged); none of the other 12 fires there.
• packages/cli unit tier, base worktree and branch: title-unresolvable 237 → 122; view-container-shape fires on its flat arm only (187, unchanged); none of the other 15 fires there.
• os validate (the built CLI) on app-crm, app-todo, app-showcase and app-multi-package, base worktree d206d08ca against the branch at 0a66efe55, and again main e540e7ffb against the final head abe4ce231, with the same result both times:
  • exit 0 each, and each prints the same 9 / 7 / 70 / 3 rule: lines;
  • app-crm, app-todo and app-multi-package are byte-identical once the worktree path and the millisecond timings are normalised;
  • app-showcase differs in exactly two lines, both of title-unresolvable on showcase_project_membership: its fix: line (237 → 122 characters of hint) and its rule: line, which gains the pointer to os explain title-unresolvable. Only this id of the 17 fires on the four apps.

Tests

Every heavy run went through scripts/pm/os-verify-lock.sh (slots issue-22161-s16, issue-22161-s16-base). Its VERDICT lines are in the report.

• Each rule's own suite pins the new shape. Each of the twelve test files gains a last block. It holds every recorded fix: line of its converted ids to one line of at most 200 characters, behind a coverage control naming each arm (every message arm of the three approval ids; each include, dimension, measure and joined-path arm of the dataset ids; both title ids on a long object name; the unknown and dotted list-view ids at both doors; all 27 hint arms of translation-target-unknown; every chart position, with and without a near miss; every object-field position; all four view-container arms; each page kind on a long page name). Exact pins read each new line, and a check per id asserts that explainRule(id) holds what the line stopped saying.
• Moved pins. Pins that read a list out of a hint moved, text-only, to the exact new line or to the verdict's "Did you mean":
  • the translation suite's list-reading pins (30), its enumeratedUniverse read-back (now the verdict's suggestion, which only a collector that reached the name can offer) and the per-app contributed-id pin (now a silent control: the contributed id resolves under its own app);
  • the list-view, object-field and chart pins on the field or measure list;
  • the approval nudge's four-policy pin and the seed pins still read the line.
• Lint suite: at the base d206d08ca Test Files 138 passed (138), Tests 6,654 passed (6,654); at 0a66efe55 138 / 6,688; after the first merge (04b1eb95b) 138 / 6,688; at the final head abe4ce231 Test Files 137 passed (137), Tests 6,654 passed (6,654), against main e540e7ffb's 137 / 6,620 (the 34 added cases, on a main feat(spec,service-automation,lint)!: a CRUD node's filter operands are value slots, the query placeholders stay — the {…} token refused (#19939 pass 4, S6) #23154 trimmed by one file).
• Lint build and typecheck: pnpm turbo run build over the CLI, metadata-protocol and example closures at 0a66efe55 (lint's build script runs check-dts-emitted) and pnpm --filter @objectstack/lint run typecheck (tsc --noEmit + check:test-typecheck OK: 2 files / 6 errors / 2 pinned signatures held), VERDICT command-exit 0 each; again at the final head abe4ce231, after pnpm install --frozen-lockfile and a rebuild of the CLI, metadata-protocol and example closures, VERDICT command-exit 0 each.
• The built entries' export closure is unchanged (the Clause-② check): the runtime export keys and the declaration exports of index, runtime and rule-explanations, read from the base dist (d206d08ca) and the branch's rebuilt dist (0a66efe55): 371 / 553, 6 / 11 and 2 / 3 names, 0 added and 0 removed. Read again at the final head abe4ce231 against a main dist built at e540e7ffb: 369 / 550, 6 / 11 and 2 / 3 names (feat(spec,service-automation,lint)!: a CRUD node's filter operands are value slots, the query placeholders stay — the {…} token refused (#19939 pass 4, S6) #23154 retired two), 0 added and 0 removed.
• CLI unit tier and metadata-protocol suite, base worktree d206d08ca against the branch at 0a66efe55, VERDICT command-exit 0 for all four runs. CLI unit tier: Test Files 285 passed (285), Tests 4,283 passed (4,283), in both. metadata-protocol: Test Files 230 passed, 3 skipped (233), Tests 28,244 passed, 19 skipped, in both. Rerun at the final head abe4ce231, because the second merge brought packages/lint changes: CLI unit tier 285 / 4,283, metadata-protocol Test Files 231 passed, 3 skipped (234), Tests 28,253 passed, 19 skipped, VERDICT command-exit 0 both, and the same per-id census as at 0a66efe55.
• Ablation: one-shot, from the committed head 515b093fd, through scripts/ablation-replace.mjs under the lock, with a shell trap restoring HEAD by absolute path. The rule suite imports the rule source, so there is no dist leg.
  • Mutation: object-field-ref-unknown's pre-slice hint restored in judge() (the prescription followed by account.detail).
  • Predicted before the run: exactly three cases in validate-object-field-refs.test.ts go red (the highlightFields exact-line pin, the relatedListColumns no-field-list pin and the new bound case), and every other case stays green.
  • Observed: Tests 3 failed, 62 passed (65), exactly those three.
  • The mutation landed (anchor x1 → x0, replacement x0 → x1, blob ed7515826362 → f3add49e4bd7) and was restored (blob ed7515826362 == HEAD, git diff HEAD empty, git status --porcelain empty).
• ESLint, narrowed: npx eslint --no-inline-config --format json over the 25 changed .ts files: 25 files, 0 errors, 0 warnings. The population is eslint.config.mjs's **/*.{ts,tsx,mts,cts,js,jsx,mjs,cjs} block plus its packages/** blocks; the config never enables type-aware linting (no parserOptions.project), so no untouched file's verdict can move. A control-character scan of every changed file found no match.
• Gates: two merges of origin/main, each through scripts/pm/os-regen-merge.sh, both without conflict; the branch edits no generated artifact.
  • 515b093fd → 04b1eb95b, onto dc93781c1: main brought packages/spec, plugin-audit and service-messaging changes and no packages/lint, packages/cli or packages/metadata-protocol file.
  • 04b1eb95b → abe4ce231, onto e540e7ffb: feat(spec,service-automation,lint)!: a CRUD node's filter operands are value slots, the query placeholders stay — the {…} token refused (#19939 pass 4, S6) #23154 ([v18] retire the {var} template dialect in flow assignment slots: refuse at registration with per-spelling remedies (the C half of #11182 ruling D, on the v18 train) #19939 S6) had landed, editing rule-explanations.ts (one flow-filter entry and its map row removed) and other packages/lint files, so every measurement above was retaken at abe4ce231.
  • node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstack, with no paths, at abe4ce231 (26 paths vs the merge base e540e7ffb) derived 62 commands, the same 62 at every head.
  • All 62 ran at abe4ce231 but the one skipped by order, each exit code captured before any pipe, all 0. The longest: check:query-options-erasure 332s, check:slot-lookup 182s, check-comment-mask-corpus 135s. At 04b1eb95b the same set ran all-0 as well, check:dual-build-cjs-loads there after a first refusal (exit 3, PREREQUISITE NOT MET: six packages had no dist) and a locked build of those six (VERDICT command-exit 0).
  • NOT MEASURED, by the dispatch's order: check:type-check-debt.
  • --ran: 62 derived famil(ies) accounted for — 61 run, 1 NOT-MEASURED.
  • origin/main moved 11 commits after the second merge (e540e7ffb … d4aaeaf0e), touching packages/lint files outside this slice (validate-component-props.ts, validate-page-field-bindings.ts, authoring-rules.ts) and none of its files or rule-explanations.ts; not merged again, and a local git merge-tree onto d4aaeaf0e is clean.
  • The artifact-roster and PR-context families are CI's.

Acceptance notes

• A long name can still lengthen a line. The bound holds on every variant the suites, the runtime probe and the four example apps fire, long names included. The names in a hint are the author's: the approver value, the object, the dataset and measure, the skill, the head segment, the translation's owning app. The closest measured case is view-container-shape's flat arm at 187 (fixed text, unchanged), then object-field-ref-unknown at 186 (the relatedListColumns prescription, fixed text).
• Some hints lose a clue a short list gave. Where no near miss exists (a field_10 in highlightFields, estimate_hours against measures named est_hours), the old line listed what exists and the new one names the remedy only. That is slices 11, 12 and 14's rule for a list the verdict's "Did you mean" covers; the two cases where the list carried the one fact an author needed (_views.default, the dead list-view tabs carrier) keep that fact as a bounded clause.
• dataset-filter-field-unknown keeps its field list. It shares checkFieldPath with dataset-field-unknown, and it is not among the claimed ids (its hint stays at most 200 on every variant the suite fires), so checkFieldPath still appends the list for it. A later slice can drop it with one condition.
• docs/adr/0079-record-display-name.md quotes the old title-format-retired hint ("titleFormat is a render-only template the server cannot return or query") as what validate-record-title.ts states. The file's module note still states that consequence, and the explanation now carries it; the ADR is a governed surface outside the claim, and the quotation records the implementation at decision time. Noted, not edited.
• Open PRs touching these files, read at the time of opening:
  • feat(spec,service-automation,lint)!: a CRUD node's filter operands are value slots, the query placeholders stay — the {…} token refused (#19939 pass 4, S6) #23154 ([v18] retire the {var} template dialect in flow assignment slots: refuse at registration with per-spelling remedies (the C half of #11182 ruling D, on the v18 train) #19939 S6) landed during this slice and is merged in (above).
  • feat(spec)!: retire the view-level api data source — a view binds its object or static rows; an API call is the page element's binding (#23082) #23160 ([Decision] Does a view's api data source adopt the page-element rule — a same-origin url and no headers written in metadata? #23082, draft, opened after the claim's census) edits one fixture of validate-list-view-field-refs.test.ts (a skip case's provider), away from this slice's hunks.

Remaining for this card

24 author-time packages/lint ids still carry a hint over 200 characters in the final lint census (at abe4ce231: 48 over 200 = 24 + 21 fenced + 3 repo-gate), by file, longest first:

• lint-flow-patterns.ts (2): flow-double-brace-interpolation 1413, flow-bare-dollar-reference 1234
• data-model-rules.ts (6): unique/legacy-organization-composite 578, unique/unscoped-declared-index 380, object/missing-name-field 327, rollup/non-numeric-aggregand 282, unique/double-declaration 279, field/select-missing-options 207
• validate-widget-bindings.ts (11): dashboard-filter-field-unprovisioned 561, chart-field-unknown 415, dashboard-filter-field-unknown 403, chart-measures-missing 369, widget-legacy-analytics-shape 338, widget-measures-missing 330, widget-filter-field-unknown 322, widget-legacy-analytics-unrenderable 320, table-count-only 311, chart-dimensions-missing 291, dashboard-filter-field-not-included 218
• validate-capability-references.ts (1): capability-reference-unknown 273
• lint-liveness-properties.ts (1): liveness-planned-property 230
• validate-page-field-bindings.ts (2): page-section-count-not-multi-value 228, page-field-unknown 206
• validate-list-view-mode.ts (1): list-view-filters-in-views-mode 219

Not in the 24: the 26 fenced ids as PR #22448 lists them (21 fire a hint over 200), the 3 repo-gate ids (startup-open-vocabulary-verdict, startup-source-unparseable, startup-verdict-assertive-wording), and the packages/cli ids on #22841.

---

Generated by Claude Code

[github-actions]

📓 Docs Drift Check

This PR changes 1 package(s): @objectstack/lint, touching 29 documentable anchor(s).

9 hand-written doc(s) NAME something this change touched and may need an implementation-accuracy re-verification:

• content/docs/api/plugin-endpoints.mdx (via /api/v1/meta/view (route, a path literal in validateViewContainers))
• content/docs/automation/approvals.mdx (via admin_rescue (literal, a string literal in validateApprovalApprovers), auto_approve (literal, a string literal in validateApprovalApprovers), org_membership_level (literal, a string literal in validateApprovalApprovers))
• content/docs/automation/flows.mdx (via auto_approve (literal, a string literal in validateApprovalApprovers))
• content/docs/concepts/metadata-lifecycle.mdx (via /api/v1/meta/view (route, a path literal in validateViewContainers))
• content/docs/getting-started/common-patterns.mdx (via auto_approve (literal, a string literal in validateApprovalApprovers))
• content/docs/kernel/services-checklist.mdx (via /api/v1/meta/view (route, a path literal in validateViewContainers))
• content/docs/plugins/packages.mdx (via org_membership_level (literal, a string literal in validateApprovalApprovers))
• content/docs/protocol/objectui/index.mdx (via /api/v1/meta/view (route, a path literal in validateViewContainers))
• content/docs/ui/forms.mdx (via /api/v1/meta/view (route, a path literal in validateViewContainers))

⛔ 5 release-owned page(s) also name something this change touched. These are read-only:

• content/docs/releases/v14.mdx (via org_membership_level (literal, a string literal in validateApprovalApprovers))
• content/docs/releases/v16.mdx (via org_membership_level (literal, a string literal in validateApprovalApprovers))
• content/docs/releases/v17/17-0.mdx (via admin_rescue (literal, a string literal in validateApprovalApprovers), auto_approve (literal, a string literal in validateApprovalApprovers))
• content/docs/releases/v17/17-6.mdx (via /api/v1/meta/view (route, a path literal in validateViewContainers))
• content/docs/releases/v17/17-7.mdx (via org_membership_level (literal, a string literal in validateApprovalApprovers))

content/docs/releases/ is RELEASE-OWNED (AGENTS.md "Documentation Guardrails"): release
notes are written centrally at release time, and a code PR that edits them is the exact PR
that guardrail exists to stop. They are still audited — read-only. If one of them is actually
wrong, file an issue or open a dedicated docs-only PR; do not edit it here.

What this run could not see

• 3 name(s) were too generic to anchor anything (single lowercase words)
• the SDK route bridge reached 54 of 206 client-bound route-ledger rows — the other 152 have no registrar path: tail to select them, so pages documenting THEIR client methods cannot appear above, on this or any run. Of those 152: 0 are remediable by widening that discovery convention (an in-repo file declares the path; the convention did not scan it); 55 are structural — on a ledger where NOT ONE row is declared in-repo, so no discovery change reaches them at any price; 97 are undecided (no in-repo declaration, on a ledger that has other in-repo registrars — absence and an unreadable spelling are not distinguishable here). The rows themselves: node scripts/docs-audit/affected-docs.mjs --bridge-coverage
• a page that states a rule by its inputs shares no identifier with the emitter that implements the rule, so an emitter-only diff cannot list it — not on this run and not on any run. Measured on fix(driver-sql): emit varchar(maxLength) for a text field a declared index keys on #11430: content/docs/protocol/objectql/types.mdx documents the text-family column mapping by the ObjectQL type names it maps FROM (text / textarea / html) while the diff changed createColumn; it went unlisted, and it was the page that diff falsified, in four places. No shared token exists to detect this on, so a rule your change carries has to be re-read by hand in the pages that restate it.
• a key NAME is not a key, so the hand re-read the line above prescribes can land on the wrong schema. The same spelling is authorable on one governed type and a [REMOVED] tombstone on another for each of active, aria, joins, objects, template, tools and version (censused on [finding] tools is a key on BOTH AgentSchema (tombstoned, dead) and SkillSchema (live, cloud-attested), so a name-based search attributes skill examples to the agent key — it produced a false stop-the-line alarm on PR #19059 #19093 over the liveness ledger's governed types, top-level keys); nothing in a search result distinguishes the two, so a grep hit on a LIVE example reads as evidence about the DEAD key. Measured on fix(spec): the agent.tools liveness row says dead — it claimed live on a key the schema tombstoned #19059: content/docs/ai/agents.mdx was reported as contradicting the agent.tools tombstone over its tools: example at :161, which is inside the defineSkill({ block opened at :155 — the page was already correct. Settle ownership by PARSING the value against both schemas, never by the name: that literal PASSES SkillSchema, and as an AgentSchema it FAILS at tools with the tombstone prescription. ⛔ These names are not the whole class — a key retired through a .strict() guidance map leaves no tombstone in the walked shape and none of them here (tool.category, live as AIToolDefinition.category).

Coarse fallback — 4 page(s) merely mention a changed package (the pre-#9192 predicate, kept for the deliberately-wide backstop): node scripts/docs-audit/affected-docs.mjs --json d4aaeaf0e2eb39ac574f1ba171a88535f9a678b4 → packageMentionDocs.

Which tree this was computed on

This run read content/docs from 4a29b554f3527aa77bbcae183ddfaf12f2687377 — the merge of head abe4ce231a02ef0248491c676b4137a940e44afe into base d4aaeaf0e2eb39ac574f1ba171a88535f9a678b4, which is what actions/checkout gives a pull_request run. Not the PR head.

A worktree cut from an older main holds a different content/docs, so re-deriving there can legitimately return a different list — that is a different tree, not a wrong row. To answer on the same tree:

# while this PR is open — GitHub drops the merge commit once it closes
git fetch origin 4a29b554f3527aa77bbcae183ddfaf12f2687377 && git checkout 4a29b554f3527aa77bbcae183ddfaf12f2687377
# afterwards, rebuild it from the two parents, which stay fetchable
git fetch origin d4aaeaf0e2eb39ac574f1ba171a88535f9a678b4 abe4ce231a02ef0248491c676b4137a940e44afe && git checkout -B drift-repro d4aaeaf0e2eb39ac574f1ba171a88535f9a678b4 && git merge --no-ff abe4ce231a02ef0248491c676b4137a940e44afe

node scripts/docs-audit/affected-docs.mjs --json d4aaeaf0e2eb39ac574f1ba171a88535f9a678b4

⚠️ That checkout carried uncommitted changes, so the commit above does not fully identify what was read.

Advisory only, and a precision-first one (#9192): a page is listed because it names a
symbol, wire route or SDK method this diff touched — not because it mentions a changed
package. Each row says which anchor put it there, so a wrong row is reportable rather than
merely annoying. To re-verify, run the docs-accuracy-audit workflow scoped to these files:
node scripts/docs-audit/affected-docs.mjs d4aaeaf0e2eb39ac574f1ba171a88535f9a678b4 → pass the list as
args.docs, on the commit named under Which tree this was computed on.