fix(types): restore the inline-locale declared face on group A's three pairs

[claude]

Fixes #9092

Group A of objectui#7759, executed against the objectui#4580 ruling already in hand (director batch #112: 「A 组按 #4580 收尾」). Three declarations restated a label key as a plain string; they now state the spec's inline locale map. packages/types/src/base.ts is untouched — it is the reference face.

Head: 52d93bc6d (patch round 1). Round-0 head: edf8b127a. Base: 2e471dc0a (origin/main at branch time, full-history clone, ⛔ not the shallow container checkout).

⭐ Patch round 1 — what the contract review changed

The ceiling-tier review (PR comment 5651847194) judged the four widenings, the ObjectViewSchema.table retirement, the ledger edits, the pin's design and the two repaired ObjectGrid sites RIGHT, and said they «should land unchanged». It failed the PR on one measured behaviour defect plus prose. Nothing in the accept set moved in this round.

#	owed	disposition
1	ObjectGrid.tsx detailTitle stringified a map label	fixed, red-first, both interpolators, string control on both legs — below
2	four «accepts the other vacuously» sentences	reworded — three of them. The fourth does not exist; measured, below
3	pin header «both are asserted here, each against its own vocabulary»	corrected, and the corrected claim is now an instrument rather than prose
4	Skill Example Check line was stale	corrected, with the mechanism — below
5	add ObjectGridSchema.title to the objectui#7759 successor list	noted under Acceptance notes. ⛔ A note, ⛔ not a fix

⭐ Item 1 — the third read site, and the reason no compiler found it

packages/plugin-grid/src/ObjectGrid.tsx, the record-detail overlay heading:

const detailTitle = schema.label
  ? t('detail.recordDetailWithLabel', { label: schema.label })
  : /* objectName branch, then the bare branch */

createSafeTranslation's t declares its options as Record-of-unknown, so widening ObjectGridSchema.label opened that arm to authors and nothing was flagged. The other two sites were found only because they are string-typed sinks and tsc named them. detailTitle feeds NavigationOverlay's title prop at three call sites, so that string IS the record-detail overlay heading — user-visible chrome, not a diagnostic.

Red-first, measured with the SAME instrument on both sides of the fix. A throwaway probe asserting the defect text was run at edf8b127a, deleted by a trap, then recreated byte-identically and run at the fixed tree:

probe case	at edf8b127a (before)	at the fixed tree (after)
provider path, map label → [object Object] Detail	passes (defect renders)	fails — text is gone
provider path, entry-less map → [object Object] Detail	passes	fails — text is gone
provider-less interpolateFallback, map label → [object Object] Detail	passes	fails — text is gone
⭐ CONTROL provider path, string label → Accounts Detail	passes	passes, unmoved
⭐ CONTROL provider-less, string label → Accounts Detail	passes	passes, unmoved

Tests 5 passed (5) before; Tests 3 failed | 2 passed (5) after, the two survivors being exactly the two string controls. The permanent pins run the other direction and were red before the fix and green after: Tests 6 failed | 5 passed (11) → Test Files 3 passed (3) · Tests 17 passed (17).

The fallthrough is placed deliberately. The resolve happens before the truthiness test, because resolveI18nLabel answers undefined for an entry-less map and '' for an empty entry — so a label that resolves to nothing lands on the objectName branch exactly as a missing label always did. Testing the raw schema.label could not: every object is truthy, and the before-probe row above confirms an entry-less map used to take the label branch and render [object Object] Detail.

:2379 columnDiagnosticLabel — re-derived, benign, CONFIRMED, ⛔ not touched

It does take the map. describeAddress in columnSpellingDiagnostics.ts gates the value on typeof label === 'string' && label.length > 0 before pushing label: '…' into its parts list, so a map-valued label is simply omitted from the message — nothing stringifies. The message is a console.warn, not chrome. The reviewer's judgment stands.

⭐ The untyped-sink sweep — what the compiler cannot do for you

Scope: reads of the four members this PR widened (AppComponentSchema.label, ObjectGridSchema.label / .description, PageNodeSchema.aria.ariaLabel), plus ObjectViewSchema.table's derived slots. Sink classes hunted by hand: t() options, String(…), template literals, JSON.stringify, Record-of-unknown parameters, any-typed props.

surface	reading	control
ObjectGrid.tsx, all schema.label / schema.description reads	four: :2379 benign-by-guard, :3219 and :4302 already resolved, :4594 fixed here. schema.description has zero reads	the same grep returns the four known schema.label lines, so the matcher fires
AppComponentSchema importers (useNavigationSync, AppSchemaRenderer, runner/App, LayoutRenderer, MetadataLoader)	no .label read on an AppComponentSchema value. The .label hits in those files are on NavigationArea / NavigationItem / action objects, and AppSchemaRenderer already routes NavigationArea.label through a resolver	AppComponentSchema itself is present in every one of the five files, so the file set is the right one
PageNodeSchema.aria	no reader in renderers/layout/page.tsx. ⚠️ Honest limit, unchanged from round 0: a grep zero cannot settle the props-spread channel. This PR widens an inert key rather than retiring one	schema.pageType in the same file, non-zero
ObjectViewSchema.table slots	plugin-view/ObjectView.tsx forwards schema.table?.description into a type: 'object-grid' object literal — a typed slot, and ObjectGrid has no description read at all. table?.label is never forwarded	table?.title in the same file, non-zero
plugin-gantt/ObjectGantt.tsx	already resolves schema.label inline; untouched	—
repo-wide String(x.label) / ${x.label} / JSON.stringify(x.label)	every hit is on option / field / view / column metadata, none on one of the four members	the sweep's own hits include the known-resolved sites, so the pattern matches

⚠️ One live defect found — filed as objectui#9373, ⛔ outside this round's fence

packages/plugin-list/src/ListView.tsx carries the identical t('detail.recordDetailWithLabel', { label: schema.label }) shape, with the same untyped sink and the same user-visible overlay heading. Measured at this branch's round-0 head with the same probe technique: map label → [object Object] Detail passes; string control → Accounts Detail passes.

⛔ It is not one of the four members this card widened — ListViewSchema.label derives from the spec's own ListViewSchema and a tsc probe shows it has admitted the map independently of this PR (control: a number into the same slot under @ts-expect-error, directive fires). So the arm was already open on main, this PR neither opens nor worsens it, and packages/plugin-list is outside this round's fence. Filed as objectui#9373 with the reproduction, the control and the repair shape; dedupe ran with a known-hit control query in the same minute.

Items 2 and 3 — the prose the pin refutes

Measured on the installed pin @objectstack/spec@17.4.0, this round, in this worktree:

• tsc: a keyed { key, defaultValue } ref into ObjectGridSchema.label / AppComponentSchema.label / PageNodeSchema.aria.ariaLabel → refused (three @ts-expect-error directives, all fire). An inline locale map into the FLAT BaseSchema.ariaLabel → refused. CONTROL: the keyed ref into that same flat slot → accepted, so the refusals are about the SLOT, not about the value.
• parse: the same keyed ref against all four mirrors → false. CONTROL: the keyed ref on the flat ariaLabel → true; the inline map there → false.
• the mechanism: InlineLocaleMapSchema types its map with key?: never; defaultValue?: never, and its own INLINE_LOCALE_KEY pattern excludes both names.

⇒ neither vocabulary admits the other, at tsc or at parse. The sentence was true when objectui#4580 Q2-B wrote it; the protocol has since closed it.

⚠️ Item 2 rests on a slightly wrong count, and here is the measurement. The review named four copies — changeset, app.ts, objectql.ts, pin header. grep -n vacuous over those four plus layout.ts returns hits in three of them: the changeset, app.ts, and the pin header. objectql.ts never carried the claim; its sentence reads "the two are structurally confusable and neither resolver accepts the other's shape" — a statement about the resolvers, and the true one, matching what layout.ts and the pre-existing base.ts already say. Control: grep -c 'confusable' is non-zero in objectql.ts, so the file was in the search set and the miss is real, not a bad path. All three that exist are reworded; ⛔ nothing was invented in objectql.ts to make the count four.

⭐ And the corrected claim is no longer prose. AGENTS.md #9 says point at the instrument rather than write the answer down — a header asserting a protocol fact is exactly the shape that rotted here. So inline-locale-declared-face-9092.test.ts gains a describe block that re-derives the cross-vocabulary refusal on both faces every run, with the control above, and the header now points at that block instead of restating it.

Item 3: the header claimed «both are asserted here, each against its own vocabulary». The flat BaseSchema.ariaLabel has no positive assertion anywhere in that file — the four widening cases are the three nested/inline members plus the BaseSchema reference face, and that reference case asserts label and description only. The header now says exactly what the file pins: about the flat key, only that an inline map is refused there.

⚠️ The header also now records the split the reviewer noted: every @ts-expect-error and typed assignment in that file is read only by tsc -p packages/types/tsconfig.test.json. vitest strips types, so a vitest-only run is a false green on that half.

Item 4 — the Skill Example Check claim was stale; ⛔ corrected

It is GREEN at this PR's head (job 103686010790: 14 of 14 ts fences judged, 0 failed). The round-0 body called it red on the strength of a branch-only local run.

⚠️ ⭐ The mechanism is the part worth keeping, because it will recur: pull_request runs check out the MERGE ref. main took objectui#9352 at 06:11Z, before this PR opened at 06:34Z, so the merge ref already carried that fix — while a branch-only local run still shows the old fence. A local check:skill-examples on a feature branch is therefore not a prediction of that check's verdict on the PR, and never was.

⚠️ Two gates ARE red, and ⛔ neither is this PR's

Both independently confirmed base-side by the review, against main's own push runs rather than against this branch:

• Doc Snippet Type Check — doc-snippet-types.yml concludes failure on main at the base 2e471dc0, at 69aa9c01, and at every push since edea22a5. The job log at this head names content/docs/guide/schema-rendering.md twice and packages/react/README.md once, all TS2740 on a DataSource shape this diff never touches. Carded objectui#9346, waiting on decision objectui#9308.
• Bundle Analysis — performance-budget.yml is success on main at 2e471dc0 and failure at 69aa9c01; the later cause is the objectui#9316 lockfile dedupe, on top of the standing ui-components debt at objectui#9204. The gate's own verdict at this head is its broken-gauge branch, with no size regression reported.

Premises, re-derived by symbol — ⛔ never by line number

Unchanged from round 0; every address was re-derived by brace-matching export interface NAME, so no reading depends on a line number the card carried.

#	premise	verdict
1	AppComponentSchema.label is plain string	CONFIRMED
2	ObjectGridSchema.label is plain string	CONFIRMED
3	ObjectGridSchema.description is plain string	CONFIRMED (it also carries @deprecated, which is not a reason to leave an authoring face narrow)
4	PageNodeSchema.aria — ⚠️ NEVER VERIFIED by anyone	CONFIRMED, and a real pair — the review re-derived it independently with six probes agreeing across three schemas
5	the card's ObjectGridSchema line numbers	REFUTED as addresses — awk-block offsets, as the filer said in place. Unused
6	producer evidence: bridgeListView assigns node.label = spec.label in react/src/spec-bridge/bridges/list-view.ts	⛔ REFUTED — the producer is GONE, retired by .changeset/retire-spec-bridge-6366.md. Independently re-derived and confirmed by the review, including the 13-site count. The ruling stands on the protocol pin and the live reads instead

The zod mirrors — the card's item 4

⭐ Every mirror already accepted the locale map, so this is declaration-only and there is no opposite-direction defect to report. app.zod.ts and objectql.zod.ts never restate the keys (so both inherit base.zod.ts's label: stripImportedDefaults(I18nLabelSchema)); layout.zod.ts receives the spec's AriaPropsSchema by reference through SpecPageFields. The narrowing was on the DECLARED side in all four cases, which is why a forward mirror-vs-declaration comparison reads these pairs as clean — zod-mirror-parity.test.ts's WiderThanDeclared ledger is the instrument that did see it, and its entries for these keys retire here. The review ablated that guard in both directions (re-adding a retired row reds; retiring a live row reds), so these retirements were required rather than chosen.

Tests

• Pin packages/types/src/__tests__/inline-locale-declared-face-9092.test.ts. Each widened key is asserted positively and paired with a @ts-expect-error negative on a genuinely-plain-string sibling (icon, objectName, ariaDescribedBy) — load-bearing, because BaseSchema carries an index signature and a key DELETED type-checks exactly like a key WIDENED. New this round: the cross-vocabulary describe block described above.
• New pin packages/plugin-grid/src/__tests__/ObjectGrid.overlayTitleInlineLocale-9092.test.tsx — the provider path: map resolved, session locale honoured (the same map picks its zh entry under a zh session), both fallthrough cases, plus the string control.
• ObjectGrid.overlayTitleNoProviderFallback.test.tsx gains the provider-less arm. It has to live there: createI18n registers its instance as react-i18next's module-global default and that registration survives cleanup(), so a no-provider render in a file that ever mounts a provider measures nothing.
• pnpm exec vitest run packages/types/ packages/plugin-grid/ → 313 files / 5409 tests passed, exit 0. Round 0 measured 312 / 5397 on the same command; the delta is +1 file and +12 tests, which is exactly what this round adds.
• pnpm --filter @object-ui/types run type-check → exit 0 (all three projects). pnpm --filter @object-ui/plugin-grid run type-check → exit 0 (tsc --noEmit && tsc -p tsconfig.test.json).
• pnpm exec turbo run build --filter="...@object-ui/types" --concurrency=2 → 43/43 tasks, exit 0. This is the real downstream check: every package's declaration build runs under dts-fail-on-type-errors.
• pnpm exec turbo run type-check --filter="...@object-ui/types" --concurrency=2 → 77/77 tasks, exit 0, grep -c 'error TS' = 0.

⚠️ The first plugin-grid type-check of this round exited 2 with a wall of TS2307: Cannot find module '@object-ui/*' — a fresh worktree with no dist/, ⛔ not a defect. Re-run after the downstream build, it is exit 0. Recorded because that failure shape reads like a regression and is not one.

Gates — exit codes captured by redirect BEFORE any pipe

All exit 0.

lint:coverage · check:entry-guard · check:upstream-port-parity · check:bash32-floor · turbo run lint --filter=@object-ui/types --filter=@object-ui/plugin-grid --force · check:vi-mock-override-shape · check:test-path-roots · check-cross-repo-closer-outcome · check:control-bytes · check:new-line-citations · check:spec-symbols · check-changeset-presence · check-changeset-no-major · check:changeset-claims

Verdict lines, read off the gates rather than restated: lint:coverage prints 46/46 packages linted, 0 with outstanding errors (0 total); check-control-bytes prints the tracked-text-file count it scanned; check-changeset-presence prints the changed/published/declared triple it compared against the merge-base.

Beyond the gate: grep -naP for control bytes over every file this round touched → no hits (exit 1).

ESLint — a narrowed run, declared as such, with the three readings the narrowing needs:

1. population: lint:coverage exit 0, verdict above — read off the gate, ⛔ not guessed;
2. this run: --format json array length 5, errorCount 0, warningCount 209 (all pre-existing; warnings are not the bar), exit 0;
3. invariance: type-aware linting is not enabled — grep -nE 'projectService|project:|tsconfigRootDir' eslint.config.js exits 1, with grep -c 'parser' = 1 as the firing control. With no type information in the lint program, this diff cannot move the verdict on any file it did not touch.

The repo-wide pnpm lint and the full pnpm test are CI's runs, ⛔ not owed here.

check-governed-queue-guard --test on this PR's paths: NOT GOVERNED — the ordinary route applies. It is a draft regardless; landing is the seat's.

Acceptance notes — ⛔ out of scope, noted

• ⭐ ObjectGridSchema.title — the objectui#7759 successor list gains this one (owed item 5). Measured on the installed pin: the protocol's ObjectGridPropsSchema.title is the string-or-inline-map union, objectui declares title?: string, and its mirror restates z.string(). So unlike this card's four keys, both objectui faces are narrow — declaration and mirror agree with each other and disagree with the protocol, the same shape as the two description notes below. ⚠️ This PR's two || schema.title fallbacks sit on it. ⛔ A note, ⛔ not a fix — it is not group A.
• AppComponentSchema.description and PageNodeSchema.description are still plain string on both faces — same shape, same successor.
• ⚠️ Candidate for the same successor list, surfaced by this round's sweep: RecordComponentAriaProps.ariaLabel (packages/types/src/record-components.ts) declares ariaLabel?: string while its docblock says it is aligned with the spec's AriaPropsSchema, whose ariaLabel is the union — and six record-component interfaces share it. ⚠️ Bounded honestly: that family has no zod mirror at all (a grep for it under packages/types/src/zod/ is empty, with the ObjectGridSchema mirror files as the firing control), so it is ⛔ not a objectui#7759 mirror-vs-declaration pair, and no author can reach the map arm through the typed face today. Noted, ⛔ not filed.
• PageNodeSchema.aria has no reader, and this PR widens an already-inert key rather than creating inertness — whether the key should exist is the ADR-0049 liveness question (objectui#4631 / objectui#7963), not this card's.
• The retired-producer citations noted under Premise 6 (base.ts, ObjectGrid.tsx, two test docblocks) are stale prose, not code. ⛔ Not swept here.

Attribution, in prose because the footer block does not reliably survive a body edit: generated with Claude Code, session session_01L5xpA5q533BgTTNADibEFt.

---

Generated by Claude Code

[github-actions]

⚠️ Console Performance Budget — gauge not trustworthy

The eager closure was measured, but one of the ceilings it is measured against no longer means what it names, so this run carries no pass/fail verdict for the performance budget.

This is not a budget violation. Nothing grew: the half marked below is a verdict about the gauge, and a ceiling that has stopped measuring anything can neither clear a bundle nor condemn one.

Step	Outcome
Build packages	success
Check console performance budget	failure

Which half objected:

Eager-closure half	Verdict
Aggregate closure ceiling	✅ pass
Per-chunk ceilings	✅ pass
Ceiling sensitivity (headroom)	⚠️ broken gauge
Ceiling freshness (checkout vs. base branch)	✅ pass

⚠️ A broken gauge half is a verdict about the ceiling, not about the bundle: that line has drifted out of range of the regression it exists to catch, or the report behind it cannot be trusted. It does not say anything grew. The Check console performance budget step log carries the ceiling and the number it was compared against.

Reason: The entry chunk measured 144.3 KB, but the eager-closure half of this gate returned no trustworthy VERDICT: the report could not be read, a ceiling has drifted out of range of the regression it must catch, or (objectui#6245) a ceiling was replaced on the base branch after this checkout was made. The step log says which. This is not a passing budget — and it is not a size regression either.

See the workflow run for details.

---

📦 Bundle Size Report

Package	Size	Gzipped
app-shell (consoleActionDispatch.js)	0.20KB	0.19KB
app-shell (index.js)	16.69KB	6.21KB
app-shell (runtime-config.js)	20.68KB	7.36KB
app-shell (types.js)	0.01KB	0.04KB
app-shell (urlParams.js)	10.06KB	3.86KB
auth (ActiveOrganizationStorage.js)	25.05KB	9.16KB
auth (AuthContext.js)	0.31KB	0.24KB
auth (AuthGuard.js)	2.07KB	1.00KB
auth (AuthProvider.js)	40.18KB	10.59KB
auth (AuthShell.js)	3.49KB	1.40KB
auth (ForgotPasswordForm.js)	12.21KB	3.45KB
auth (LoginForm.js)	18.15KB	5.39KB
auth (PreviewBanner.js)	0.90KB	0.50KB
auth (RegisterForm.js)	6.65KB	2.22KB
auth (SocialSignInButtons.js)	9.61KB	3.89KB
auth (UserMenu.js)	3.41KB	1.23KB
auth (auth-gate-events.js)	1.29KB	0.66KB
auth (authStyles.js)	5.04KB	1.72KB
auth (createAuthClient.js)	40.21KB	10.80KB
auth (createAuthenticatedFetch.js)	8.46KB	3.43KB
auth (index.js)	3.19KB	1.44KB
auth (invitation-status.js)	1.22KB	0.70KB
auth (org-roles.js)	6.66KB	2.78KB
auth (phone-identifier.js)	1.11KB	0.66KB
auth (types.js)	0.59KB	0.35KB
auth (useAuth.js)	5.30KB	1.02KB
auth (useWorkspaceAdminStatus.js)	11.08KB	4.58KB
collaboration (CommentThread.js)	26.08KB	7.56KB
collaboration (LiveCursors.js)	3.17KB	1.27KB
collaboration (PresenceAvatars.js)	6.49KB	2.64KB
collaboration (PresenceProvider.js)	2.79KB	1.13KB
collaboration (index.js)	1.68KB	0.73KB
collaboration (useCollaborationTranslation.js)	6.05KB	2.52KB
collaboration (useCommentSearch.js)	1.98KB	0.88KB
collaboration (useConflictResolution.js)	7.75KB	1.86KB
collaboration (useMentionNotifications.js)	1.81KB	0.68KB
collaboration (usePresence.js)	6.33KB	1.84KB
collaboration (useRealtimeSubscription.js)	7.91KB	2.01KB
components (index.js)	502.02KB	115.16KB
core (index.js)	8.52KB	3.41KB
create-plugin (index.js)	27.94KB	9.51KB
data-objectstack (index.js)	211.58KB	58.68KB
fields (index.js)	247.89KB	62.50KB
i18n (LocalizationContext.js)	1.76KB	0.96KB
i18n (builtinAggregateLabels.js)	0.86KB	0.49KB
i18n (currency.js)	1.22KB	0.64KB
i18n (fallbackInterpolation.js)	6.25KB	2.77KB
i18n (i18n.js)	8.87KB	3.64KB
i18n (index.js)	5.22KB	2.26KB
i18n (pickLocalized.js)	9.86KB	3.95KB
i18n (provider.js)	32.15KB	10.49KB
i18n (useDisplayLocale.js)	2.85KB	1.45KB
i18n (useObjectLabel.js)	34.34KB	9.17KB
i18n (useSafeTranslation.js)	5.60KB	2.33KB
layout (index.js)	38.83KB	10.95KB
mobile (MobileProvider.js)	0.92KB	0.49KB
mobile (ResponsiveContainer.js)	0.94KB	0.38KB
mobile (breakpoints.js)	1.51KB	0.70KB
mobile (createOfflineDataSource.js)	5.61KB	1.75KB
mobile (index.js)	1.99KB	0.87KB
mobile (offlineQueue.js)	3.91KB	1.35KB
mobile (pwa.js)	0.97KB	0.49KB
mobile (serviceWorker.js)	1.48KB	0.62KB
mobile (serviceWorkerSource.js)	3.41KB	1.48KB
mobile (useBreakpoint.js)	1.54KB	0.65KB
mobile (useGesture.js)	6.96KB	1.98KB
mobile (useOfflineSync.js)	1.99KB	0.72KB
mobile (usePullToRefresh.js)	2.53KB	0.85KB
mobile (useResponsive.js)	0.72KB	0.42KB
mobile (useSpecGesture.js)	4.39KB	1.66KB
mobile (useTouchTarget.js)	1.01KB	0.54KB
permissions (MePermissionsProvider.js)	13.52KB	4.88KB
permissions (PermissionContext.js)	0.31KB	0.25KB
permissions (PermissionGuard.js)	0.89KB	0.45KB
permissions (PermissionProvider.js)	6.24KB	2.16KB
permissions (discardProofCache.js)	1.04KB	0.55KB
permissions (evaluator.js)	8.39KB	3.10KB
permissions (index.js)	0.93KB	0.41KB
permissions (store.js)	0.91KB	0.42KB
permissions (useFieldPermissions.js)	1.28KB	0.53KB
permissions (usePermissions.js)	4.83KB	2.27KB
plugin-ai (index.js)	14.81KB	3.63KB
plugin-calendar (index.js)	49.25KB	13.99KB
plugin-charts (index.js)	71.51KB	19.97KB
plugin-chatbot (index.js)	195.34KB	46.51KB
plugin-dashboard (index.js)	131.22KB	34.59KB
plugin-designer (index.js)	215.94KB	44.33KB
plugin-detail (index.js)	253.46KB	65.85KB
plugin-editor (index.js)	2.23KB	1.05KB
plugin-form (index.js)	136.77KB	34.17KB
plugin-gantt (index.js)	166.95KB	41.04KB
plugin-grid (index.js)	211.65KB	57.50KB
plugin-kanban (index.js)	46.01KB	14.30KB
plugin-list (index.js)	112.58KB	27.65KB
plugin-map (index.js)	20.64KB	6.86KB
plugin-markdown (index.js)	13.88KB	4.80KB
plugin-report (index.js)	43.41KB	11.93KB
plugin-timeline (index.js)	30.07KB	8.74KB
plugin-tree (index.js)	9.55KB	3.32KB
plugin-view (index.js)	84.42KB	20.79KB
providers (DataSourceProvider.js)	0.75KB	0.39KB
providers (MetadataProvider.js)	1.37KB	0.59KB
providers (ThemeProvider.js)	1.90KB	0.85KB
providers (UploadProvider.js)	11.66KB	3.50KB
providers (index.js)	0.45KB	0.23KB
providers (types.js)	0.01KB	0.04KB
react-runtime (index.js)	5.62KB	2.34KB
react (LazyPluginLoader.js)	4.47KB	1.63KB
react (SchemaRenderer.js)	94.03KB	31.02KB
react (data-invalidation.js)	5.05KB	2.08KB
react (index.js)	4.63KB	2.18KB
react (schema-input.js)	4.25KB	2.04KB
react (spec-input.js)	0.20KB	0.18KB
sdui-parser (codegen.js)	6.58KB	2.74KB
sdui-parser (dashboard-widget-options.js)	3.08KB	1.30KB
sdui-parser (index.js)	5.66KB	2.50KB
sdui-parser (input-type.js)	2.84KB	1.40KB
sdui-parser (kanban-quick-add.js)	3.89KB	1.87KB
sdui-parser (parse.js)	25.28KB	7.80KB
sdui-parser (provenance.js)	3.66KB	1.82KB
sdui-parser (types.js)	0.28KB	0.23KB
sdui-parser (validate.js)	14.82KB	4.99KB
types (ai.js)	0.20KB	0.17KB
types (api-types.js)	0.20KB	0.18KB
types (app.js)	2.87KB	1.00KB
types (base.js)	0.20KB	0.18KB
types (blocks.js)	0.20KB	0.18KB
types (complex.js)	2.93KB	1.49KB
types (crud.js)	0.20KB	0.18KB
types (dashboard-filter-alias.js)	6.23KB	2.74KB
types (data-display.js)	3.75KB	1.85KB
types (data-protocol.js)	0.20KB	0.19KB
types (data.js)	0.20KB	0.18KB
types (designer.js)	1.85KB	0.85KB
types (disclosure.js)	0.20KB	0.18KB
types (error-code.js)	1.54KB	0.88KB
types (expression.js)	0.20KB	0.18KB
types (feedback.js)	0.20KB	0.18KB
types (field-types.js)	0.20KB	0.18KB
types (form.js)	0.20KB	0.18KB
types (http-inflight.js)	8.87KB	3.73KB
types (http-retry.js)	4.32KB	2.02KB
types (icon-key-migration.js)	4.26KB	1.63KB
types (index.js)	4.74KB	2.25KB
types (layout.js)	0.20KB	0.18KB
types (managed-by.js)	0.19KB	0.18KB
types (mobile.js)	4.73KB	2.28KB
types (navigation.js)	0.20KB	0.18KB
types (objectql.js)	0.20KB	0.18KB
types (overlay.js)	0.20KB	0.18KB
types (permissions.js)	0.20KB	0.18KB
types (plugin-scope.js)	0.20KB	0.18KB
types (record-components.js)	0.20KB	0.19KB
types (record-semantics.js)	1.28KB	0.67KB
types (registry.js)	0.20KB	0.18KB
types (reports.js)	0.20KB	0.18KB
types (select-option.js)	0.20KB	0.19KB
types (spec-report.js)	5.05KB	1.93KB
types (spec-ui-namespace.js)	0.20KB	0.19KB
types (strict-authoring-face.js)	14.27KB	5.47KB
types (system-fields.js)	3.33KB	1.54KB
types (theme.js)	6.28KB	2.87KB
types (ui-action.js)	8.11KB	3.32KB
types (views.js)	0.20KB	0.18KB
types (widget.js)	0.20KB	0.18KB

Size Limits

• ✅ Core packages should be < 50KB gzipped
• ✅ Component packages should be < 100KB gzipped
• ⚠️ Plugin packages should be < 150KB gzipped

[os-sam]

CI red at edf8b127 — established as NOT this PR's, from the run's own log. Standing down, third PR in this lane.

Doc Snippet Type Check (job 103686010833) exit 1. Read from the job log, ⛔ not from the PR author's local re-run and ⛔ not from an aggregate conclusion:

[semantic] content/docs/guide/schema-rendering.md:99:29   TS2740: … missing … from type 'DataSource<any>': find, findOne, create, update, and 2 more.
[semantic] content/docs/guide/schema-rendering.md:434:29  TS2740: … same
[semantic] packages/react/README.md:73:29                 TS2740: … same
Semantic phase: 649 of 649 block(s) judged, 3 failed.

The instrument was live — its own controls fired

⛔ A red is as void as a zero if nothing proves the harness ran. This gate prints its controls, and all four discriminate:

control	reading
resolution	@object-ui/types → packages/types/dist/index.d.ts
sentinel (must fail)	ThisNameIsDefinitelyNotExported → 1 diagnostic (TS2305)
positive (must pass)	BaseSchema → 0 diagnostics
undeclared (must fail)	@floating-ui/react-dom → 1 (TS2307)
root-declared (must fail)	vitest → 1 (TS2307)

⇒ 649 of 649 blocks reached the semantic phase and exactly 3 failed. A real reading.

Disjoint from this PR's file face

This PR changes seven files: .changeset/9092-inline-locale-declared-face.md · packages/plugin-grid/src/ObjectGrid.tsx · packages/types/src/app.ts · packages/types/src/layout.ts · packages/types/src/objectql.ts · packages/types/src/__tests__/inline-locale-declared-face-9092.test.ts · packages/types/src/__tests__/zod-mirror-parity.test.ts.

⇒ Neither content/docs/guide/schema-rendering.md nor packages/react/README.md is among them, and the diagnostic subject — DataSource<any>'s six members — is a shape this diff never mentions. The widening here is string → string | I18nLabel on four declaration members.

⭐ Three PRs, three disjoint diffs, one identical failure

This is now the decisive reading, and it is stronger than it was an hour ago:

PR	its diff	these two checks
objectui#9349	packages/types/src/zod/** + a pin	❌ same 3 + 1
objectui#9353	one file, content/docs/api/schema-reference.md	❌ same 3 + 1
objectui#9364 (this)	packages/types decls + plugin-grid	❌ same 3 + 1

Three PRs whose file faces do not intersect cannot each independently break the same three fences. ⇒ the failure is on the base, ⛔ not on any of them.

Root cause is carded, and ⛔ there is no fix in existence to port

The objectui#7912 / PR objectui#9310 renderer-seam narrowing meeting fences already on main:

half	card	state
the three doc fences	objectui#9346	domain:ui · p2 · pm:blocked
the skills fence (Skill Example Check)	objectui#9311 → PR objectui#9352	in flight

objectui#9346 establishes all three routes to green are refused by something already landed — a cast at the seam re-creates the defect one line lower (objectui#7912's own execution note 2); widening the declaration back is the card (decision batch #68, maintainer 「同意」); and declaring the blocks in the gate's ledger is a quarantine, not a reason. What remains is rewriting what those pages teach, which is objectui#9308's decision, ⛔ not a documentation edit — and schema-rendering.md is itself pinned to the wiring it teaches today.

⇒ ⛔ nothing to port. The cross-lane direct-takeover route needs the blocking item to be mechanical, clearly specified and S-sized; this one is none of the three — it waits on a maintainer decision.

⛔ What was not done

⛔ No re-run. A deterministic TS2740 against content this PR does not touch cannot come back different; the one permitted re-run is preserved for a case where it could discriminate. ⛔ No test skipped, disabled or quarantined. ⛔ No gate weakened. ⛔ No empty commit. ⛔ No fourth card — both halves are carded and dedupe hit.

State

Stays draft. ⛔ Not flipped ready, ⛔ not enqueued — this lane's bar is every check green, ⛔ not the required subset. Its other 34 checks are green or running clean, and the Clause-②: yes ceiling-tier contract review is the next gate regardless.

⚠️ Recorded from objectui#9346, ⛔ not re-measured here: Doc Snippet Type Check is reportedly not in main's required set. That would mean it blocks nothing in the merge queue — but it does not lift this lane's stricter bar, so it changes nothing above.

Seat: domain:spec @ objectui, os-sam, session session_01L5xpA5q533BgTTNADibEFt, 2026-09-13T06:42Z.

---

Generated by Claude Code

[os-sam]

Bundle Analysis red at edf8b127 — taken to the chunk's own card, as the gate directs

Second red on this PR, from a different base-side debt than the doc fences in comment 5651716838.

The gate's instruction, followed literally:

Paying the row down is the open decision on the chunk, ⛔ not a task for whichever change the queue happened to weigh — take it there, and say on this pull request that you did.

⇒ Done: objectui#9204 comment 5651743555.

What failed

Three of the four eager-closure halves passed; the objecting half is ceiling sensitivity on the ui-components per-chunk row, which the gate reports as "ALREADY declared exhausted before this run" and as "a standing debt that predates this change … measured at three gzipped bytes across five unrelated merges."

⚠️ Measured vs. inferred, stated precisely. This seat read the per-chunk block from objectui#9349's job log (ui-components 387.8 KB / 389.6 KB, headroom 1.9 KB = 0.02x, under the 0.10x floor). For this PR's run it read only the rendered verdict table — aggregate ✅, per-chunk ✅, sensitivity ⚠️ broken gauge, freshness ✅, entry chunk 144.3 KB — and ⛔ did not open the per-chunk block. So the mechanism is established and the specific ui-components figure for this run is ⛔ NOT MEASURED here.

✅ Freshness passed ⇒ ⛔ not the objectui#6245 stale-ceiling shape.

⛔ Nothing done that would turn it green: ⛔ never raise the ceiling, ⛔ never raise the allowance, ⛔ no re-run. This PR's diff is packages/types declarations plus two plugin-grid read sites; it touches no packages/components file.

State

Stays draft. This PR now carries two reds, neither of which it is permitted to fix: the DataSource doc fences (objectui#9346 → decision objectui#9308) and this chunk row (objectui#9204). Its ceiling-tier contract review is in flight and is a separate gate from both.

Seat: domain:spec @ objectui, os-sam, session session_01L5xpA5q533BgTTNADibEFt, 2026-09-13T06:51Z.

---

Generated by Claude Code

[os-sam]

Contract review

Head reviewed: edf8b127ad3def8326be8df5d8ae724179fc0c90 — the current PR head, unmoved.
Tier: ceiling (claude-fable-5-1), context-isolated adversarial subagent — fed the card, the standing rulings and the PR only; ⛔ never the dispatch order or this seat's conclusions, and told in as many words that this seat's card comments are not evidence and that two of its instructions on the sibling card had already been refuted by measurement.

Tier — MEASURED, ⛔ not self-reported. Transcript grepped on the strict harness-stamp form "model":"…": 115 stamps, 115 × claude-fable-5-1, 0 non-fable. Control fires. ⭐ And this transcript contains no bare non-fable token at all — the self-seeding hazard recorded earlier (a prior record quoting its own tier, fetched back by the next reviewer) did not arise here.

Independence pair (machine-readable)

Implemented-by: claude/issue-9092-group-a-inline-locale
Reviewed-by: session_01L5xpA5q533BgTTNADibEFt

⛔ Not SELF-REVIEW.

⛔ VERDICT: FAIL — and ⛔ this one is NOT prose-only

Unlike the sibling card's rounds, this FAIL carries a measured behaviour defect. The four declaration widenings, the table retirement, the ledger edits, the pin's design and the two repaired grid sites are all judged RIGHT and land unchanged. But the read-site audit missed a third site on the same key in the same file:

ObjectGrid.tsx:4578 passes schema.label straight into t('detail.recordDetailWithLabel', { label: schema.label }). Because t's options are Record<string, unknown>, the compiler could not name it — so widening the declaration opened that arm to authors and left it rendering "[object Object] Detail", measured on both the i18next provider path and the provider-less interpolateFallback path, against a string control returning "Accounts Detail". It feeds NavigationOverlay title= at three sites: the record-detail overlay title, user-visible.

⇒ The card's own mechanic — every read of a widened key must resolve — is not yet satisfied. Patch round 1 of 2 is dispatched.

⛔ A correction to this seat's own record on this PR

The ruling caught an error of mine. My comment 5651716838 asserted this PR failed «the same 3 + 1» — three doc fences plus one skill fence. Re-read from the check runs just now: Skill Example Check (job 103686010790) concludes success. 36 checks, exactly two failures: Doc Snippet Type Check and Bundle Analysis.

⚠️ The cause is worth carrying forward, because it will bite again: pull_request runs check out the MERGE ref. main took objectui#9352 (28be0786) at 06:11Z, before this PR opened at 06:34Z, so its merge ref already carries that fix — while a dev's branch-only local run still shows the old fence. I took the author's local reading and my own pattern from a sibling PR instead of reading this PR's job. ⛔ That is the same fault I have been correcting in others all round.

⚠️ NOT re-measured yet: whether the +1 is likewise wrong on objectui#9349 and objectui#9367. Both heads postdate 06:11Z, so the same mechanism probably applies, but ⛔ probably is not a reading — I am re-reading both and will correct those comments separately rather than let a second wrong claim stand.

⭐ The ruling also independently confirmed the two reds that are real, on the base rather than on this PR — doc-snippet-types.yml concludes failure on main at the base 2e471dc0, at 69aa9c01, and at every push since edea22a5; performance-budget.yml is success at 2e471dc0 and failure at 69aa9c01 (objectui#9316, lockfile dedupe). ⇒ the Bundle Analysis red has a second, later cause than the standing ui-components debt I recorded at objectui#9204; both readings are consistent, and the base-side attribution holds either way.

Carriers

Both needs:contract-review carriers — this PR and card objectui#9092 — are stripped in the same stroke as this record, per 「FAIL 同 PASS 剥双载体」. ⛔ Stripping is not a clearance: the gate is cleared only by the round-1 record, and the carriers are re-hung the moment the patch head lands. Card state and assignee unchanged.

---

Ruling — verbatim, ⛔ unedited

⛔ The text below is the isolated ceiling-tier subagent's ruling as returned. This seat's only legal actions are verbatim adoption or wholesale voiding; it is ⛔ never rewritten, abridged or polished. The framing above is this seat's, ⛔ not a substitute for it.

---

Contract review

Head reviewed: edf8b127ad3def8326be8df5d8ae724179fc0c90 (draft PR #9364, card objectui#9092, group A of objectui#7759).

Measured in a private linked worktree cut from refs/pull/9364/head, with pnpm install --frozen-lockfile; base pinned by commit as 2e471dc0 (the PR's stated branch point, fetched into a review-owned ref, not read from a remote-tracking name), current main read as 69aa9c01. Every origin/main reading below is by content. Protocol read from the installed pin @objectstack/spec@17.4.0, which is also the single @objectstack/spec@ entry in pnpm-lock.yaml and the ^17.4.0 floor in packages/types/package.json. Every ablation below was proven mutated on disk (grep count and git hash-object) before its instrument ran, and proven restored by git diff --quiet empty and blob hash equal to HEAD:<path>; every exit code was captured by redirect before any pipe. The worktree and the three review refs were removed afterwards.

① Derived judgments — every accept-set and public-face move in the diff

Seven files, +274/−41, exactly as declared. In the three declaration files nothing moves but the four members below, two import type additions and docblocks.

1. AppComponentSchema.label: string → string | I18nLabel — RIGHT. Protocol: the spec's AppSchema.label is ZodUnion<[ZodString, InlineLocaleMap]> (dist/app.zod-C1Ice8CN.d.ts, AppSchema block, member 23). Mirror: zod/app.zod.ts:377 is BaseSchema.extend(SpecAppFields.shape).extend({…}) with 'label' in its except list (:365) and no restatement, so it inherits base.zod.ts:303 stripImportedDefaults(I18nLabelSchema). The declared set now equals the reference face BaseSchema.label (objectui#4580 revised Q1, option A) and the protocol's own arm; nothing is admitted the mirror refuses (see item 6 for the one spec-owned residual). Ablation A1 (re-narrow to string): tsc -p packages/types/tsconfig.test.json exit 2 — the new pin fires TS2322 at inline-locale-declared-face-9092.test.ts(50,56) AND the ledger fires twice, WiderLedgerMismatch naming "app.zod.ts#AppComponentSchema" and WiderLedgerKeyDrift naming "label". Ablation A2 (delete the member outright): exit 0 — correctly, because deletion falls back to the identical inherited BaseSchema.label; the accept set does not move, so a green is a reading, not a vacuity (this is proposal A1's own spelling on objectui#7759).
2. ObjectGridSchema.label: string → string | I18nLabel — RIGHT. Mirror zod/objectql.zod.ts:234 is BaseSchema.extend({…}) restating neither label nor description. Protocol: ObjectGridPropsSchema.label on the pin is the same union (dist/ui/index.d.ts, ObjectGridPropsSchema block, member 3).
3. ObjectGridSchema.description: string → string | I18nLabel — RIGHT. Same inheritance (base.zod.ts:310); reference face BaseSchema.description. @deprecated kept; deprecated-but-declared is still an authoring face and the mirror already admitted the map.
4. PageNodeSchema.aria.ariaLabel: string → string | I18nLabel — RIGHT, and it is a real pair. Mirror zod/layout.zod.ts:684 is BaseSchema.extend(SpecPageFields.shape), and SpecPageFields (:517) excludes only name, label, description, type, regions — aria passes through from the spec's PageSchema, whose top-level aria is AriaPropsSchema with ariaLabel: ZodOptional<ZodUnion<[ZodString, InlineLocaleMap]>> (dist/ui/index.d.ts:454). By-reference verified behaviourally, six probes agreeing across PageNodeSchema mirror, spec AriaPropsSchema and spec PageSchema.aria: inline map true; keyed {key, defaultValue} false; keyed with params false; { english } false; extra sibling key false (the spec's strictObject, so the nested object is strict inside a passthrough BaseSchema); number false. Ablation A3 (delete only the ariaLabel line): exit 2 — the pin fires TS2353/TS2339 at (76,61)/(84,26) and two standing p1-spec-alignment.test.ts pins fire with it. Ablation A4 (delete the whole aria member): exit 2, exactly one error, TS2578: Unused '@ts-expect-error' directive at (81,5) — the deletion type-checked through BaseSchema's [key: string]: any (base.ts:492) precisely as the PR reasoned, and the paired negative caught it.
5. ObjectViewSchema.table ledger entry retired — RIGHT. objectql.ts:1920 declares table?: Partial<Pick<ObjectGridSchema, ObjectGridSlotKey>>; ObjectGridSlotKey (:1719–1790) contains both 'label' and 'description' (each counted once); the mirror is ObjectGridSchema.omit({type, objectName}).partial() (objectql.zod.ts:399). The repair reaches it mechanically.
6. Bounded by the protocol, with one spec-owned residual. I18nLabel on the pin is z.input<typeof I18nLabelSchema> = string | (Record<string,string> & { key?: never; defaultValue?: never }) (dist/ui/index.d.ts:796–859). tsc admits { english: 'x' } (probe line 9, no diagnostic) where every mirror refuses it (INLINE_LOCALE_KEY regex, dist/ui/index.js:2630; measured false on AppComponentSchema.label, ObjectGridSchema.description, PageNodeSchema.aria.ariaLabel; { en } controls true). That gap is the spec's own type-vs-regex distance, identical on the reference face BaseSchema, and not this PR's to close.
7. WiderThanDeclared ledger and WIDER_ARMS — RIGHT. Five rows leave (App::label, Page::aria, Grid::label, Grid::description, View::table); header 22/35/45 — 6/29/0/10 → 20/30/37 — 5/25/0/7 re-derives (−2 entries; −5 keys; −8 arms = 2+1+2+2+1; −1 SCHEMA-NODE = table; −4 CONCRETE; −3 unions = the three two-arm label/description rows). Both files green on the head (vitest run 2 files / 38 tests, exit 0). Ablation A5a (re-add the retired ObjectGridSchema entry): exit 2, WiderLedgerMismatch names "objectql.zod.ts#ObjectGridSchema". Ablation A5b (retire the LIVE ChatbotSchema::body from ledger and WIDER_ARMS): exit 2, names "complex.zod.ts#ChatbotSchema" and "body". The guard fires in both directions, so the retirements here were required, not chosen.
8. ObjectGrid.tsx:3219 (export viewLabel) and :4302 (data-table caption) — RIGHT on both arms. resolveI18nLabel (dist/ui/index.js:3252) returns a string unchanged, undefined for null/undefined, and for a map exact → base → regional → default → en → first string entry, undefined only for an entry-less map; readEntry returns '' for an empty entry, so '' || schema.title falls to title exactly as the string arm always did. displayLocale is in scope (:1154); buildExportFileName.viewLabel?: string (core/src/utils/export-filename.ts:50) is now satisfied. Pre-PR harm re-derived: caption reaches data-table.tsx:1972 {caption && <TableCaption>{caption}</TableCaption>} — an object there is a React throw, which is what "reached the caption as an object" means; the filename path is [object Object]. Reachable for any map label the mirror already accepted.
9. WRONG — the read-site audit is incomplete on the same key in the same file. ObjectGrid.tsx:4578–4579: schema.label ? t('detail.recordDetailWithLabel', { label: schema.label }) : …. createSafeTranslation's t takes options?: Record<string, unknown> (i18n/src/useSafeTranslation.ts:26), so the compiler could not name this site and the PR's inventory did not reach it. Measured: on the provider path (i18next 26.4.0, '{{label}} Detail') the map renders "[object Object] Detail"; on the provider-less path interpolateFallback (fallbackInterpolation.ts:103, String(v)) renders "[object Object] Detail"; string control "Accounts Detail". detailTitle feeds NavigationOverlay title= at :4904, :5199, :5258 — the record-detail overlay title, user-visible. This is the t() interpolation site objectui#7759's proposal A (comment 5596479714) priced into group A by name, and objectui#4580's mechanics require each read to resolve. The PR opens this arm to tsc authors and leaves it rendering [object Object]. (:2379 columnDiagnosticLabel also takes the map; describeAddress drops non-strings, so the diagnostic merely loses its label — benign.)
10. Non-regression outside the four members. No other declaration, key or accept set moves. AppComponentSchema importers (app-shell/hooks/useNavigationSync.ts, layout/AppSchemaRenderer.tsx, runner/App.tsx, runner/LayoutRenderer.tsx, runner/lib/MetadataLoader.ts) read no .label on it; ObjectViewSchema.table?.description (plugin-view/ObjectView.tsx:1707) lands in a typed slot; ObjectGantt.tsx:1950 already resolves inline. PageNodeSchema.aria has no reader in renderers/layout/page.tsx (schema.aria 0, control schema.pageType 2) — a grep zero cannot settle the props-spread channel (AGENTS.md), but this widens an inert key rather than retiring one. CI on the merge ref (head + 69aa9c01): Type Check, Build & E2E, Lint and all four Test shards success; 36 check runs reported = 36 returned (30 + 6, two pages).

② Semver grading

Changeset declares @object-ui/types: minor, @object-ui/plugin-grid: patch. Consistent. A widening of a published type surface is source-breaking for a downstream TS reader of these members (this PR's own two TS2322s in plugin-grid are the proof) and this repository grades that minor with the semantics spelled in the body — precedent .changeset/hidden-predicate-widen-7455.md says so in its last paragraph, and objectui#4581's BaseSchema widening shipped under ## 17.5.0 / ### Minor Changes (packages/types/CHANGELOG.md:465–492). plugin-grid patch matches a runtime repair on the map arm with the string arm unchanged. Advisory only: the body never says in one sentence what a downstream reader of the four members must now do (resolve through resolveI18nLabel); the plugin-grid paragraph implies it. The body also carries one measured-false sentence — ③ below — and it publishes verbatim into the CHANGELOG.

③ Boundary flags — each answered or escalated

• Premise 6, retired producer. Re-derived on the head tree: git ls-tree -r refs/pull/9364/head -- packages/react/src/spec-bridge/ → 0 entries; git grep -c bridgeListView -- packages/react/src → 0 files; control SchemaRenderer.tsx → 20 hits. The "13 live inline-resolver source sites" claim re-counted = 13. Refutation confirmed; the ruling stands on the pin and the live reads. Seat took option B (objectui#9365) — agreed; nothing further.
• Skill Example Check declared red — it is GREEN at this head. Job 103686010790: Semantic phase: 14 of 14 ts fence(s) judged, 0 failed, all four controls fired. Cause: pull_request runs check out the merge ref, and main had already taken fix(skills): guard the DataSource read in the marked data-integration example #9352 (28be0786, 06:11Z) before the PR opened (06:34Z); the branch tree alone still carries the old fence, which is what the author's local run and the seat's PR comment 5651716838 ("same 3 + 1") measured. Not a defect; the record on the PR is stale on this line.
• Doc Snippet Type Check red, declared not this PR's — CONFIRMED on base, independently. doc-snippet-types.yml push runs on main conclude failure at 2e471dc0 (the base) and at 69aa9c01, and at every push since edea22a5 (fix(react)!: dataSource at the renderer seam is the published DataSource contract #9310); the job log at this head names content/docs/guide/schema-rendering.md:99,434 and packages/react/README.md:73, all TS2740 on DataSource<any> — no file or shape in this diff.
• Bundle Analysis red, not flagged by the PR. The PR read CI while this job was in progress. Verdict at head is the gate's broken-gauge branch (bot comment 5651710231: aggregate, per-chunk and freshness pass; sensitivity "broken gauge"; no size regression). performance-budget.yml on main: success at 2e471dc0, failure at 69aa9c01 (chore(deps): run pnpm dedupe --lockfile-only on an untouched main — the measurement (objectui#9215) #9316, lockfile dedupe) — base-side. Not this PR's.
• "The two object shapes … each accept the other vacuously" — measured-FALSE on the installed pin, stated in four new places (changeset, app.ts docblock, objectql.ts docblock, pin header). On 17.4.0 InlineLocaleMapSchema's type carries key?: never; defaultValue?: never and INLINE_LOCALE_KEY refuses both names at parse. tsc probe: keyed ref into I18nLabel → TS2322 twice; inline map into KeyedI18nLabel → TS2353; keyed into BaseSchema.label, inline into flat BaseSchema.ariaLabel, keyed into nested aria.ariaLabel — all refused; control (number into I18nLabel) refused. Zod: I18nLabelSchema keyed → false; AriaPropsSchema keyed → false. The sentence was true at objectui#4580 Q2-B's time and is quoted from there; the pin has since closed it. Effect on the new pin's controls: none — its negatives use string-typed siblings (icon, objectName, ariaDescribedBy), never the other vocabulary. Effect on the tree: four copies of a claim an instrument now refutes, one CHANGELOG-bound. Must be reworded.
• Pin header: "both are asserted here, each against its own vocabulary." The flat BaseSchema.ariaLabel is asserted nowhere in the file (the fourth case asserts label/description only). Prose inaccuracy; fix in the same round.
• The two-sided pin reasoning (question 7) — sound where it matters. BaseSchema's index signature does make a deleted member type-check like a widened one; the paired @ts-expect-error catches that for the only member with no inherited twin (aria, A4), and for label/description a deletion is a no-op on the accept set (A2), so the pin's green there is correct. The type half of this pin is read only by tsc -p packages/types/tsconfig.test.json (the package type-check script and CI's Type Check job), not by vitest — the pin header does not say so; the parity file's own guards say it about themselves.
• Ledger header edit vs AGENTS.md 完善设计器的每一个细节 #9 — compliant. The figures the PR moved are pinned live figures: the test at zod-mirror-parity.test.ts:4309 reads them back by regex every run and compares them to ledgerEntryKeys, WIDER_ARMS and the mirrors. Ablation A6 (**20 entries** → **21 entries**, prose only): vitest exit 1, expected { entries: 21 … } to deeply equal { entries: 20 … }. A figure an instrument re-derives on every run is the rule's "point at the instrument", not the rule's "write down its answer"; a stale figure cannot survive a run here. Not owed: the "It read … until objectui#NNNN" history sentence earlier retirements added — no pin reads it.
• open_questions Q2 (commit trailer) — the seat answered A; outside this review's remit; noted.
• Acceptance notes verified: AppComponentSchema.description and PageNodeSchema.description narrow on both faces (app.zod.ts:381, layout.zod.ts PageNode extend body description: z.string()); successor objectui#7759. One to add to that successor's list, same family, not group A: ObjectGridSchema.title — the protocol's ObjectGridPropsSchema.title is the union, objectui declares title?: string and its mirror restates z.string() (map probe false); this PR's two new || schema.title fallbacks sit on it.

Verdict

FAIL. The four accept-set moves, the table retirement, the ledger edits, the pin's design and the two repaired grid sites are right and should land unchanged. Two things must move first, in one round: (a) ObjectGrid.tsx:4578 — resolve schema.label through resolveInlineI18nLabel(schema.label, displayLocale) before the t() call (keeping the objectName fallthrough when the map resolves to nothing), with a red-first pin that a map label yields Accounts Detail, not [object Object] Detail, and a string control that does not move; (b) reword the four "accepts the other vacuously" sentences to what the pin measures (neither vocabulary admits the other on 17.4.0, at tsc and at parse — cite InlineLocaleMapSchema's key?: never), and the pin header's "both are asserted here". The PR body's Skill Example Check line is stale and may be corrected in passing. No accept-set change is asked for.

Not reached

• Emitted dist byte-identity for @object-ui/types: reasoned from the two additions being import type (no JS emitted), not measured by building base and head.
• The SchemaRenderer props-spread channel for PageNodeSchema.aria: grep zero with a firing control only; no runtime probe, since this PR widens an inert key rather than retiring one.
• Repo-wide turbo build / type-check in the review worktree: not run; CI's Type Check, Build & E2E and four test shards on the merge ref were taken as the downstream instrument.
• The doc-snippet and bundle gates were not run locally; main's own push runs at the pinned base and current head were taken instead.
• Note for the seat: fetching refs/heads/main into a review-owned ref opportunistically advanced the shared checkout's origin/main 2e471dc0 → 69aa9c01 (the remote's true state; no working-tree change). Shared HEAD untouched at 4f2d1f4a.

Implemented-by: claude/issue-9092-group-a-inline-locale
Reviewed-by: session_01L5xpA5q533BgTTNADibEFt

---

Generated by Claude Code

[github-actions]

⚠️ Console Performance Budget — gauge not trustworthy

The eager closure was measured, but one of the ceilings it is measured against no longer means what it names, so this run carries no pass/fail verdict for the performance budget.

This is not a budget violation. Nothing grew: the half marked below is a verdict about the gauge, and a ceiling that has stopped measuring anything can neither clear a bundle nor condemn one.

Step	Outcome
Build packages	success
Check console performance budget	failure

Which half objected:

Eager-closure half	Verdict
Aggregate closure ceiling	✅ pass
Per-chunk ceilings	✅ pass
Ceiling sensitivity (headroom)	⚠️ broken gauge
Ceiling freshness (checkout vs. base branch)	✅ pass

⚠️ A broken gauge half is a verdict about the ceiling, not about the bundle: that line has drifted out of range of the regression it exists to catch, or the report behind it cannot be trusted. It does not say anything grew. The Check console performance budget step log carries the ceiling and the number it was compared against.

Reason: The entry chunk measured 144.3 KB, but the eager-closure half of this gate returned no trustworthy VERDICT: the report could not be read, a ceiling has drifted out of range of the regression it must catch, or (objectui#6245) a ceiling was replaced on the base branch after this checkout was made. The step log says which. This is not a passing budget — and it is not a size regression either.

See the workflow run for details.

---

📦 Bundle Size Report

Package	Size	Gzipped
app-shell (consoleActionDispatch.js)	0.20KB	0.19KB
app-shell (index.js)	16.69KB	6.21KB
app-shell (runtime-config.js)	20.68KB	7.36KB
app-shell (types.js)	0.01KB	0.04KB
app-shell (urlParams.js)	10.06KB	3.86KB
auth (ActiveOrganizationStorage.js)	25.05KB	9.16KB
auth (AuthContext.js)	0.31KB	0.24KB
auth (AuthGuard.js)	2.07KB	1.00KB
auth (AuthProvider.js)	40.18KB	10.59KB
auth (AuthShell.js)	3.49KB	1.40KB
auth (ForgotPasswordForm.js)	12.21KB	3.45KB
auth (LoginForm.js)	18.15KB	5.39KB
auth (PreviewBanner.js)	0.90KB	0.50KB
auth (RegisterForm.js)	6.65KB	2.22KB
auth (SocialSignInButtons.js)	9.61KB	3.89KB
auth (UserMenu.js)	3.41KB	1.23KB
auth (auth-gate-events.js)	1.29KB	0.66KB
auth (authStyles.js)	5.04KB	1.72KB
auth (createAuthClient.js)	40.21KB	10.80KB
auth (createAuthenticatedFetch.js)	8.46KB	3.43KB
auth (index.js)	3.19KB	1.44KB
auth (invitation-status.js)	1.22KB	0.70KB
auth (org-roles.js)	6.66KB	2.78KB
auth (phone-identifier.js)	1.11KB	0.66KB
auth (types.js)	0.59KB	0.35KB
auth (useAuth.js)	5.30KB	1.02KB
auth (useWorkspaceAdminStatus.js)	11.08KB	4.58KB
collaboration (CommentThread.js)	26.08KB	7.56KB
collaboration (LiveCursors.js)	3.17KB	1.27KB
collaboration (PresenceAvatars.js)	6.49KB	2.64KB
collaboration (PresenceProvider.js)	2.79KB	1.13KB
collaboration (index.js)	1.68KB	0.73KB
collaboration (useCollaborationTranslation.js)	6.05KB	2.52KB
collaboration (useCommentSearch.js)	1.98KB	0.88KB
collaboration (useConflictResolution.js)	7.75KB	1.86KB
collaboration (useMentionNotifications.js)	1.81KB	0.68KB
collaboration (usePresence.js)	6.33KB	1.84KB
collaboration (useRealtimeSubscription.js)	7.91KB	2.01KB
components (index.js)	502.02KB	115.16KB
core (index.js)	8.52KB	3.41KB
create-plugin (index.js)	27.94KB	9.51KB
data-objectstack (index.js)	211.58KB	58.68KB
fields (index.js)	247.89KB	62.50KB
i18n (LocalizationContext.js)	1.76KB	0.96KB
i18n (builtinAggregateLabels.js)	0.86KB	0.49KB
i18n (currency.js)	1.22KB	0.64KB
i18n (fallbackInterpolation.js)	6.25KB	2.77KB
i18n (i18n.js)	8.87KB	3.64KB
i18n (index.js)	5.22KB	2.26KB
i18n (pickLocalized.js)	9.86KB	3.95KB
i18n (provider.js)	32.15KB	10.49KB
i18n (useDisplayLocale.js)	2.85KB	1.45KB
i18n (useObjectLabel.js)	34.34KB	9.17KB
i18n (useSafeTranslation.js)	5.60KB	2.33KB
layout (index.js)	38.83KB	10.95KB
mobile (MobileProvider.js)	0.92KB	0.49KB
mobile (ResponsiveContainer.js)	0.94KB	0.38KB
mobile (breakpoints.js)	1.51KB	0.70KB
mobile (createOfflineDataSource.js)	5.61KB	1.75KB
mobile (index.js)	1.99KB	0.87KB
mobile (offlineQueue.js)	3.91KB	1.35KB
mobile (pwa.js)	0.97KB	0.49KB
mobile (serviceWorker.js)	1.48KB	0.62KB
mobile (serviceWorkerSource.js)	3.41KB	1.48KB
mobile (useBreakpoint.js)	1.54KB	0.65KB
mobile (useGesture.js)	6.96KB	1.98KB
mobile (useOfflineSync.js)	1.99KB	0.72KB
mobile (usePullToRefresh.js)	2.53KB	0.85KB
mobile (useResponsive.js)	0.72KB	0.42KB
mobile (useSpecGesture.js)	4.39KB	1.66KB
mobile (useTouchTarget.js)	1.01KB	0.54KB
permissions (MePermissionsProvider.js)	13.52KB	4.88KB
permissions (PermissionContext.js)	0.31KB	0.25KB
permissions (PermissionGuard.js)	0.89KB	0.45KB
permissions (PermissionProvider.js)	6.24KB	2.16KB
permissions (discardProofCache.js)	1.04KB	0.55KB
permissions (evaluator.js)	8.39KB	3.10KB
permissions (index.js)	0.93KB	0.41KB
permissions (store.js)	0.91KB	0.42KB
permissions (useFieldPermissions.js)	1.28KB	0.53KB
permissions (usePermissions.js)	4.83KB	2.27KB
plugin-ai (index.js)	14.81KB	3.63KB
plugin-calendar (index.js)	49.25KB	13.99KB
plugin-charts (index.js)	71.51KB	19.97KB
plugin-chatbot (index.js)	195.34KB	46.51KB
plugin-dashboard (index.js)	131.22KB	34.59KB
plugin-designer (index.js)	215.94KB	44.33KB
plugin-detail (index.js)	253.46KB	65.85KB
plugin-editor (index.js)	2.23KB	1.05KB
plugin-form (index.js)	136.77KB	34.17KB
plugin-gantt (index.js)	166.95KB	41.04KB
plugin-grid (index.js)	211.66KB	57.50KB
plugin-kanban (index.js)	46.01KB	14.30KB
plugin-list (index.js)	112.58KB	27.65KB
plugin-map (index.js)	20.64KB	6.86KB
plugin-markdown (index.js)	13.88KB	4.80KB
plugin-report (index.js)	43.41KB	11.93KB
plugin-timeline (index.js)	30.07KB	8.74KB
plugin-tree (index.js)	9.55KB	3.32KB
plugin-view (index.js)	84.42KB	20.79KB
providers (DataSourceProvider.js)	0.75KB	0.39KB
providers (MetadataProvider.js)	1.37KB	0.59KB
providers (ThemeProvider.js)	1.90KB	0.85KB
providers (UploadProvider.js)	11.66KB	3.50KB
providers (index.js)	0.45KB	0.23KB
providers (types.js)	0.01KB	0.04KB
react-runtime (index.js)	5.62KB	2.34KB
react (LazyPluginLoader.js)	4.47KB	1.63KB
react (SchemaRenderer.js)	94.03KB	31.02KB
react (data-invalidation.js)	5.05KB	2.08KB
react (index.js)	4.63KB	2.18KB
react (schema-input.js)	4.25KB	2.04KB
react (spec-input.js)	0.20KB	0.18KB
sdui-parser (codegen.js)	6.58KB	2.74KB
sdui-parser (dashboard-widget-options.js)	3.08KB	1.30KB
sdui-parser (index.js)	5.66KB	2.50KB
sdui-parser (input-type.js)	2.84KB	1.40KB
sdui-parser (kanban-quick-add.js)	3.89KB	1.87KB
sdui-parser (parse.js)	25.28KB	7.80KB
sdui-parser (provenance.js)	3.66KB	1.82KB
sdui-parser (types.js)	0.28KB	0.23KB
sdui-parser (validate.js)	14.82KB	4.99KB
types (ai.js)	0.20KB	0.17KB
types (api-types.js)	0.20KB	0.18KB
types (app.js)	2.87KB	1.00KB
types (base.js)	0.20KB	0.18KB
types (blocks.js)	0.20KB	0.18KB
types (complex.js)	2.93KB	1.49KB
types (crud.js)	0.20KB	0.18KB
types (dashboard-filter-alias.js)	6.23KB	2.74KB
types (data-display.js)	3.75KB	1.85KB
types (data-protocol.js)	0.20KB	0.19KB
types (data.js)	0.20KB	0.18KB
types (designer.js)	1.85KB	0.85KB
types (disclosure.js)	0.20KB	0.18KB
types (error-code.js)	1.54KB	0.88KB
types (expression.js)	0.20KB	0.18KB
types (feedback.js)	0.20KB	0.18KB
types (field-types.js)	0.20KB	0.18KB
types (form.js)	0.20KB	0.18KB
types (http-inflight.js)	8.87KB	3.73KB
types (http-retry.js)	4.32KB	2.02KB
types (icon-key-migration.js)	4.26KB	1.63KB
types (index.js)	4.74KB	2.25KB
types (layout.js)	0.20KB	0.18KB
types (managed-by.js)	0.19KB	0.18KB
types (mobile.js)	4.73KB	2.28KB
types (navigation.js)	0.20KB	0.18KB
types (objectql.js)	0.20KB	0.18KB
types (overlay.js)	0.20KB	0.18KB
types (permissions.js)	0.20KB	0.18KB
types (plugin-scope.js)	0.20KB	0.18KB
types (record-components.js)	0.20KB	0.19KB
types (record-semantics.js)	1.28KB	0.67KB
types (registry.js)	0.20KB	0.18KB
types (reports.js)	0.20KB	0.18KB
types (select-option.js)	0.20KB	0.19KB
types (spec-report.js)	5.05KB	1.93KB
types (spec-ui-namespace.js)	0.20KB	0.19KB
types (strict-authoring-face.js)	14.27KB	5.47KB
types (system-fields.js)	3.33KB	1.54KB
types (theme.js)	6.28KB	2.87KB
types (ui-action.js)	8.11KB	3.32KB
types (views.js)	0.20KB	0.18KB
types (widget.js)	0.20KB	0.18KB

Size Limits

• ✅ Core packages should be < 50KB gzipped
• ✅ Component packages should be < 100KB gzipped
• ⚠️ Plugin packages should be < 150KB gzipped

[os-sam]

Contract review

Head reviewed: 52d93bc6d6dd20a7c9f0dffb146520916b11a651 (previous head edf8b127, FAILED at record 5651847194).
Round: narrow re-review of patch round 1 — scoped to what that FAIL owed, plus whether this round disturbed what it judged right.
Tier: ceiling (claude-fable-5-1), context-isolated. MEASURED, strict harness-stamp form: 54 stamps, 54 fable, 0 non-fable, control fires. No self-seeded token.

Independence pair (machine-readable)

Implemented-by: claude/issue-9092-group-a-inline-locale
Reviewed-by: session_01L5xpA5q533BgTTNADibEFt

⛔ Not SELF-REVIEW.

⭐ VERDICT: PASS

⭐ The discriminating reading is ablation A-naive, and it is the one that matters. Reverting the fix reddens six cases with [object Object] Detail three times in the failure DOM — right reason, not timing. But the reviewer also ran the plausible wrong fix: moving the resolve inside the branch instead of before the truthiness test. That reddens exactly the three fallthrough pins (entry-less map on both interpolators, empty entry on the provider path), with [object Object] absent from the log.

⇒ the pins separate the deliberate fix from a naive one that would have satisfied every map case and still been wrong. That is the difference between a pin that catches the defect and a pin that catches the class, and neither I nor the FAIL asked for it.

⚠️ And the prior ruling was wrong on one count — confirmed independently. Its owed item 2 named four copies of the «accepts the other vacuously» sentence. Re-derived at the reviewed head with both grep -c (lines) and grep -o | wc -l (occurrences) agreeing: changeset 1, app.ts 1, pin header 1, objectql.ts 0, layout.ts 0 — with grep -c confusable = 1 in objectql.ts as the same-subject control proving the file was in the search set, and a tree-wide git grep vacuous confirming it never carried the word. Three existed, three were reworded. The author declined to invent a fourth and was right to.

⭐ The corrected prose is now true rather than less false, and rests on a measurement re-derived here: resolveI18nLabel({ key: 'a.b', defaultValue: 'D' }, 'en') → "a.b" via its final any-string-entry loop. The resolver does not refuse the other vocabulary — it answers wrongly. The type and parse faces are what refuse (three @ts-expect-error directives firing; safeParse false on both mirrors, map controls true).

⭐ The new instrument block discriminates on both runners, proven by three separate ablations (C1 widening the grid member → TS2578 at the exact directive; C2 widening the flat BaseSchema.ariaLabel → TS2578 at the block's refused directive; C3 widening the mirror → the parse assertion red). The runner split the FAIL flagged is real and now disclosed in the header: the type half is read only by tsc -p packages/types/tsconfig.test.json, ⛔ never by vitest.

Nothing else moved, by blob hash: six paths in r1..r2; objectql.ts, layout.ts, base.ts, zod-mirror-parity.test.ts (which holds both the ledger and WIDER_ARMS) and all three zod mirrors are blob-identical to the previous head and to base. The app.ts hunk is 13 changed lines, every one a * docblock line, zero non-comment lines.

objectui#9373 (the ListView twin) is confirmed the right home: that shape is present at the base too, the arm was open independently of this card's four members, and this PR touches zero files under packages/plugin-list.

Carriers — cleared, ⛔ not stripped

Both limbs, one stroke each, per 「清标即落地:PASS ⇒ 同席同笔剥双载体」. Record id and judged head are in this comment.

⛔ Landing — ③ still fails

① PASS on record at tier ✅ · ② dual carrier cleared, machine-read ✅ · ③ every check green ❌. CI at this head: 36 reported = 36 returned, all four Test shards + Type Check + Lint + Build & E2E + Skill Example Check + Governed Surface Queue Guard success; exactly two failures, Doc Snippet Type Check and Bundle Analysis, both base-side. ⛔ Not ready, ⛔ not enqueued — blocked only by objectui#9308 (via objectui#9346) and objectui#9204 / objectui#9316.

③ Q1 answered — flag 2 gets its own card, ⛔ not a line on objectui#7759

The reviewer left one advisory question to this seat: whether base.ts:466-479 — still saying «each accepts the other's shape vacuously», the one surviving copy in packages/types/src of a sentence this PR's new instrument now refutes — belongs on objectui#7759's successor list or on its own card.

Its own card: objectui#9374. Reasons: it sits on the reference face every widened member inherits from, ⛔ not on a narrow-declaration pair, so it is a different kind of thing from the two description notes already on that list; objectui#7759 is a large multi-part family and a note there waits on the whole family; and ⭐ an instrument that refutes it now exists in the tree, which makes it cheaply checkable rather than a prose opinion. ⛔ Correctly out of scope here — this card forbids touching BaseSchema.

The other two advisories are recorded and ⛔ carry no round: objectql.ts:682-683 («neither resolver accepts the other's shape» — true of the signatures, misleading at runtime; «neither resolver's signature accepts» would close it), and the changeset's «An earlier draft of this note said …» parenthetical, which publishes into the CHANGELOG where no reader saw the earlier draft. Both ride any future push that touches those files.

---

Ruling — verbatim, ⛔ unedited

⛔ The isolated ceiling-tier subagent's ruling as returned. Verbatim adoption or wholesale voiding are this seat's only legal actions.

---

Contract review

Head reviewed: 52d93bc6d6dd20a7c9f0dffb146520916b11a651 (patch round 1 of draft PR #9364, card objectui#9092); previous head edf8b127ad3def8326be8df5d8ae724179fc0c90, FAILED by record 5651847194. Narrow re-review: only what that FAIL owed, plus whether this round disturbed what it judged right.

Measured in a private linked worktree (/home/user/objectui-review-9364-r2, removed afterwards) at a review-owned ref fetched by sha, pnpm install --frozen-lockfile exit 0, installed pin @objectstack/spec@17.4.0 (matches the ^17.4.0 floor in packages/types/package.json). The shared checkout was never edited or stashed; its HEAD is 4f2d1f4a before and after. Every ablation was proven mutated on disk (grep count and git hash-object differing from HEAD:<path>) before its instrument ran, and proven restored by blob hash equal to HEAD:<path> and git diff --quiet; the worktree was git status-clean at teardown. Every exit code was captured by redirect before any pipe.

① Derived judgments

A. The behaviour defect is fixed, and the pins discriminate — including the half a naive fix gets wrong.

1. The site. ObjectGrid.tsx:4594–4596 now reads const resolvedDetailLabel = resolveInlineI18nLabel(schema.label, displayLocale); const detailTitle = resolvedDetailLabel ? t('detail.recordDetailWithLabel', { label: resolvedDetailLabel }) : <objectName branch> : <bare branch>. resolveInlineI18nLabel is the spec's resolveI18nLabel under its objectui alias (:48), displayLocale is useDisplayLocale() at :1154, the same pair the two round-0 repairs at :3219 and :4302 use. detailTitle still feeds NavigationOverlay title= at :4921, :5216, :5275. The resolve sits before the truthiness test, as the FAIL asked.
2. The fallthrough premise, re-derived on the pin rather than read from the comment: resolveI18nLabel({}, 'en') → undefined; resolveI18nLabel({ en: '' }, 'en') → ''; string → unchanged; undefined → undefined; { zh: 'X' } under en → 'X' (first-entry arm). So an entry-less map and an empty entry are both falsy and land on the objectName branch. The old schema.label ? … was truthy for every object — the FAIL's mechanism, confirmed.
3. Baseline at head: pnpm exec vitest run on the two plugin-grid pin files plus the types pin → Test Files 3 passed (3) · Tests 21 passed (21), exit 0.
4. Ablation A-revert (site restored to the round-0 form, label: schema.label, resolvedDetailLabel count 0 on disk): vitest exit 1, Tests 6 failed | 5 passed (11). The six reds are exactly the map cases — provider-path map, provider-path zh session, provider-path entry-less map, provider-path empty entry, provider-less map, provider-less entry-less map. The failure DOM carries [object Object] Detail three times, so the reds are for the RIGHT reason. The five survivors are the two string controls (Accounts Detail, unmoved on both interpolators) plus the three pre-existing no-provider tests. Restored; blob equals HEAD:packages/plugin-grid/src/ObjectGrid.tsx (d6bab0c5).
5. Ablation A-naive (resolve moved INSIDE the branch: schema.label ? t(…, { label: resolveInlineI18nLabel(schema.label, displayLocale) })): vitest exit 1, Tests 3 failed | 8 passed (11) — precisely the three fallthrough pins (entry-less map on both interpolators, empty entry on the provider path), with [object Object] absent from the log. This is the discriminating reading: the pins separate the deliberate before-the-test resolve from the naive fix that would have satisfied the map cases alone. Restored by blob hash.
6. The provider-less arm lives in ObjectGrid.overlayTitleNoProviderFallback.test.tsx, where the file-split rule (react-i18next's module-global default surviving cleanup()) makes it measurable; the author's reason for the placement holds and the ablation shows that arm reddening independently (interpolateFallback's String(v) at fallbackInterpolation.ts:103).

B. The count was three, not four — the prior ruling was wrong on it, and the author is right. Re-derived at edf8b127 (the head the FAIL reviewed), with grep -c (lines) and grep -o | wc -l (occurrences) agreeing: changeset 1, app.ts 1, pin header 1, objectql.ts 0, layout.ts 0. Same-subject control: grep -c confusable in objectql.ts = 1, so the file was in the search set and the miss is real. A tree-wide git grep vacuous at that ref returns 100+ hits and confirms objectql.ts never carried the word. The objectql.ts sentence is «the two are structurally confusable and neither resolver accepts the other's shape», which is a different claim (flag ③.1). Declining to invent a fourth copy was correct.

The reworded prose rests on a measurement I re-derived on the pin: resolveI18nLabel({ key: 'a.b', defaultValue: 'D' }, 'en') → "a.b" (also "a.b" without defaultValue, and under zh); the pin's source at dist/ui/index.js:3252 reaches it through the final «any string entry» loop. So the resolver does not refuse the other vocabulary, it answers wrongly — the author's claim is TRUE. Parse face: I18nLabelSchema.safeParse(keyed) → false, AriaPropsSchema.safeParse({ ariaLabel: keyed }) → false, map controls true; INLINE_LOCALE_KEY at dist/ui/index.js:2630 excludes key/defaultValue by name. Type face: the three @ts-expect-error directives in the new block all fire at head (tsc -p packages/types/tsconfig.test.json exit 0, 0 errors) — a keyed ref is refused on every widened member and an inline map on the flat ariaLabel, with the keyed ref accepted there as the control. Every sentence the three rewordings and the corrected header now make — «refused at tsc and at parse alike», «resolveI18nLabel hands a keyed reference back as its own key string», «BaseSchema.ariaLabel has no positive assertion anywhere in this file» — is true against the installed pin and the file as it stands. The header now also records the runner split the FAIL noted.

C. The new instrument block discriminates on both runners.

• tsc half, C1: widening ObjectGridSchema.label (objectql.ts:685) to also admit { key; defaultValue?; params? } → tsc exit 2, inline-locale-declared-face-9092.test.ts(159,5): TS2578 Unused '@ts-expect-error' — the grid directive in the new block, plus one zod-mirror-parity error and one in objectql.ts (3 total). Restored by blob hash.
• tsc half, C2: widening the flat BaseSchema.ariaLabel (base.ts:486) to also admit I18nLabel → exit 2, (172,5): TS2578 — the block's refused directive — plus the standing base-schema-label-vocabulary pin and parity (3 total). Restored by blob hash.
• parse half, C3: widening the mirror's label (base.zod.ts:303) to a union admitting the keyed ref → vitest exit 1, Tests 1 failed | 9 passed (10), the one red being parse: the mirrors refuse a keyed ref on the same keys that take the map (expected true to be false). Restored by blob hash.
• The runner split is real and disclosed: the type half is read only by tsc -p packages/types/tsconfig.test.json (include: src/**/*.test.ts, chained from the package type-check script); vitest strips types. Each half has a firing control inside the block (keyed ref accepted on the flat slot; inline map refused there), so neither passes on a slot that refuses every object.

D. Nothing else moved — by blob hash. git diff --name-status r1..r2 lists exactly six paths: the changeset, ObjectGrid.tsx, the new grid pin, the no-provider pin, the types pin, app.ts. Blob identity r1 == r2 for objectql.ts (6f252e76), layout.ts (6742e366), base.ts (f9d9…→ identical to base too), zod-mirror-parity.test.ts (30ad6fd2, which holds both the WiderThanDeclared ledger and WIDER_ARMS), and all three zod mirrors (app.zod.ts, objectql.zod.ts, layout.zod.ts, each identical to base as well). The app.ts r1→r2 hunk is 13 changed lines, every one a * docblock line, zero non-comment lines changed; label?: string | I18nLabel; is at the same two lines (407, 416) in both. The accept set, the four widenings, the table retirement and the ledger are exactly what the FAIL judged right.

E. objectui#9373 — filed, not fixed — is the right disposition. plugin-list/src/ListView.tsx:2387–2388 carries the identical t('detail.recordDetailWithLabel', { label: schema.label }) shape at this head AND at the base 2e471dc0 (line 2388); this PR touches zero files under packages/plugin-list. A tsc probe against the head tree: an inline map assigned to ListViewSchema['label'] compiles, with a @ts-expect-error on a number in the same slot firing as the control (exit 0, probe file removed). ListViewSchema is z.input of the zod ListViewSchema extending BaseSchema, whose label is the spec's I18nLabelSchema — the arm was open before this card, independent of the four widened members. The card is group A's three declaration pairs; the sibling defect is a READ on a fifth, un-widened member in a package outside the fence, so the fix belongs to its own card. objectui#9373 exists (07:52Z, open), carries reproduction, control, the repair shape with the before-the-test resolve, and names #9364 with no closing keyword. Correct.

CI at this head: 36 check runs reported = 36 returned. All four Test shards, Type Check, Lint, Build & E2E, Skill Example Check and Governed Surface Queue Guard success. Exactly two failures — Doc Snippet Type Check and Bundle Analysis — the base-side debts the prior ruling confirmed on main's own push runs; not the contract question, not re-derived this round.

② Semver grading

Changeset unchanged in kind: @object-ui/types: minor, @object-ui/plugin-grid: patch. Consistent, and unchanged from the prior ruling's grading. The third repaired site is a runtime repair on the map arm of a member the mirror already admitted, with the string arm measured byte-identical on both interpolators (the controls) — patch for plugin-grid. check-changeset-presence exit 0 (9 files changed, 8 published source of 2 released packages, 1 changeset declared); check-changeset-no-major exit 0. The body's new paragraph correctly names the untyped-sink class (t() options, String(…), template literals, JSON.stringify) as the thing a widening's compile errors cannot reach; a downstream reader of the four members has the instruction it needed.

③ Boundary flags and open_questions

1. objectql.ts:682–683 «neither resolver accepts the other's shape» — true on the signature, misleading at runtime; advisory. resolveI18nLabel(label: I18nLabel | undefined, …) and resolveKeyedI18nLabel(label: string | { key: string; … }) (app-shell/src/utils/index.ts) each refuse the other vocabulary at tsc. At runtime the first returns 'a.b' for a keyed ref — this PR's own new sentence in three other places. The word «accepts» is doing the work; «neither resolver's signature accepts» would close the gap. Round-0 prose the FAIL did not name; not blocking.
2. base.ts:466–479 still says «each accepts the other's shape vacuously». Pre-existing on main (blob identical base == head), the card forbids touching BaseSchema, and the FAIL did not name it. It is now the one remaining copy in packages/types/src of a sentence the new instrument block refutes, two properties from the reference face. Belongs on a follow-up note against objectui#4580 Q2-B's docblocks, not on this PR.
3. Changeset parenthetical «An earlier draft of this note said …» publishes into the CHANGELOG, where no reader saw the earlier draft. Harmless noise; the substantive sentence it qualifies is true.
4. Runner split on the plugin-grid pins: their GridLabel-typed assignments are read by tsc -p packages/plugin-grid/tsconfig.test.json, which needs the downstream dist; CI Type Check success at this head is the instrument for that half (see Not reached).

open_questions: none blocking. Q1 (advisory): whether flag 2 should be a note on objectui#7759's successor list or a one-line docblock card of its own — the seat's call.

Not reached

• tsc -p packages/plugin-grid/tsconfig.test.json locally (fresh worktree, no downstream dist); CI Type Check success at 52d93bc6 taken. Does not change the verdict — the runtime half of those pins was ablated here on both interpolators.
• vitest under the C1/C2 type-face mutations, to show the vitest-only false green directly: reasoned from vitest stripping types, not measured. Does not change the verdict; the tsc reds at the exact directive lines are the reading that matters.
• Full pnpm test, turbo build, doc-snippet and bundle gates: CI's four shards, Build & E2E and the base-side attribution from the prior ruling taken; not re-run.

Verdict

PASS. The owed behaviour defect is fixed at the site the FAIL named, red-first on both interpolators with string controls that do not move, and the fallthrough pins separate the deliberate fix from the naive one. The «four copies» premise in the prior ruling was wrong: three existed, three were reworded, and the corrected claim is true on the installed pin and is now re-derived by an instrument that reddens on both its runners when either vocabulary leaks. Nothing outside the six round-1 paths moved, by blob hash. objectui#9373 is the right home for the ListView twin.

Implemented-by: claude/issue-9092-group-a-inline-locale
Reviewed-by: session_01L5xpA5q533BgTTNADibEFt

---

Generated by Claude Code

[os-sam]

⛔ Correction to the record above — the follow-up card is objectui#9375, ⛔ not objectui#9374

Record 5652151842 (the PASS above) answers the reviewer's Q1 with «Its own card: objectui#9374.» ⛔ Wrong number. The card was filed after that comment was written and is objectui#9375. ⛔ objectui#9374 is not this card and nothing in that record refers to it.

Everything else in the record stands: the verdict, the ruling, the carrier disposition, and the reasoning for filing a separate card rather than a note on objectui#7759.

⛔ How it happened, since the mechanism is the point

I wrote a predicted issue id into a permanent record instead of filing first and quoting the real one. ⛔ That is a named prohibition in this seat's own discipline — «⛔ Never write a predicted issue or comment id» — recorded earlier in this very session after the same mistake was caught before posting. This time it reached a posted record.

⚠️ It is also the same failure class this round has been about from both ends: a claim written as fact that no instrument had re-derived. The id was available for the cost of one call, in the right order, and I did not pay it.

⇒ Standing, and this correction is the second entry against it: ⛔ file first, then quote the id the API returned. ⛔ No exception for "it will obviously be the next number."

---

Generated by Claude Code

[os-sam]

⛔ Correction: a fix for the Doc Snippet Type Check half now EXISTS — objectui#9369. Still not portable.

My standing-down comment 5651716838 said the chain had «⛔ no fix in existence to port». True when written, now false.

objectui#9369 carries Fixes #9308 and Fixes #9346, on the maintainer's 2026-09-13 ruling (verbatim 「同意」, option B), and is not draft. It moves the exact three DataSource fences that fail on this head.

⛔ Why it is still not ported here

1. The doc fix depends on the code change. The fences move to PredicateScopeProvider, a wiring that exists only after objectui#9369's b1/b2 land. Porting the fence edits alone teaches a provider this tree does not have.
2. Porting the whole PR is not "the same change" — 35 files, +889/−282, with a declared breaking flip of data.* visibility predicates. ⛔ Not a rider on a declaration-widening PR.
3. This PR's fence is packages/types + packages/plugin-grid; content/docs/** and packages/react/** are outside it.

⇒ disposition unchanged: wait for objectui#9369. ⛔ Still draft, ⛔ still not enqueued.

⚠️ The second blocker is independent and untouched: Bundle Analysis's exhausted ui-components row (objectui#9204 / objectui#9316). Both must clear.

Contract state complete: ceiling-tier PASS at 52d93bc6 (record 5652151842), both carriers cleared, --pair 9364 exit 0.

---

Generated by Claude Code

[os-sam]

CI status — both reds are INHERITED, and one of them is provably stale. ⛔ Not pushing yet.

domain:spec @ objectui seat (session_01L5xpA5q533BgTTNADibEFt), 2026-09-13T09:5xZ.

Doc Snippet Type Check fails on three fences — content/docs/guide/schema-rendering.md:99 and :434, and packages/react/README.md:73 — which are exactly the three PR #9369 named and repaired. #9369 merged 2026-09-13T09:02:22Z (main 85243729) and main is now green on this check (its own run, job 103701523599 at 09:06Z, leaves only Bundle Analysis red). This PR's diff touches none of those files, and its failing run was created 2026-09-13T07:51:46Z, 1h11m before #9369 landed.

Bundle Analysis is main's own standing debt — objectui#9204, dispatched and in flight.

⛔ The re-run was spent on the sibling PR as a probe, and it does not help

rerun_failed_jobs on PR #9349's run 34743258393 produced attempt 2 at 09:42:41Z and failed with the identical three fences ⇒ a pull_request re-run reuses the merge commit recorded at first trigger rather than recomputing it against the current base. The fix reaches a PR only through a new merge ref, i.e. a push. ⇒ re-running here would be a known no-op and is deliberately not done.

The sequencing

check-clause2-carriers.mjs --pair 9364 reads exit 0 with a review of record naming this head. Merging main in now would move the head — stalling that review and owing a re-hang of both carriers plus a fresh review — while leaving Bundle Analysis red anyway.

⇒ one base merge after objectui#9204 lands and main is green on both rows. This PR stays watched until then. ⛔ Nothing here is a verdict on its diff.

---

Generated by Claude Code