chore(objectui): bump the console pin to 87af769e9a3e with 27 re-measured citations, and fold in the two-code sdui-parser port (#17645)

[os-project-manager]

Fixes #17429
Part of #17645 — that card's deliverable (the two-code port) lands here; the director seat ends card 17645 with the landing record after the merge (a seat-written record, not a closing reference). The card's take-over from the offline domain:devx seat is the maintainer's instruction (「18723 os-try-charles 已下班，你也接手」), recorded on the card; the carrier reader's cross-author row (C9) cannot be cleared without the offline holder's own Release:, so the card is carried as Part of, not as a closing reference.

Clause-②: yes

Corrected from no under ruling batch #201 item 1 (maintainer 「201 A」, record 5753263180 on #17645): this PR now also carries #17645's two-code sdui-parser port, so the save gate's diagnostic set grows (inert-quick-add, member-type-mismatch); the pin bump and the 27 re-measured citations alone moved no accept set. Both landed on this branch: the fold-in commit 6ef8344645 and the citation-spelling fix fc25a5925d (the head); the lockstep record needed no re-recording (see the fold-in section).

(Pin-bump half, as originally declared:) Declared from the diff, not from the card: nothing in the pin bump adds, removes or renames an authorable key. The two packages/spec edits with a published face are .describe() SENTENCES (Dataset measure format, FormField.span); check:authorable-surface and check:api-surface both report their artefacts current with no regeneration, and pnpm --filter @objectstack/spec check:generated reports all 15 artefacts up to date after gen:docs and gen:migration-registry ran. The same authored metadata is accepted and refused as before, so neither arm of the closed pair applies.

What moved

.objectui-sha: 53ded82bf7a494f54e344e19099dbf00854b8694 → 87af769e9a3ee28ace099fdd653d3ebd79fe82e2 (objectui origin/main tip read 2026-09-20; cfcc17d9dd04 — the revision PR #18723 ported from — is an ancestor of it, git merge-base --is-ancestor exit 0).

Produced by scripts/bump-objectui.sh --no-commit 87af769e9a3e, never by hand. Everything the pin travels with:

Artefact	How
.changeset/console-87af769e9a3e.md	the bump script's own digest — 584 releasing changesets of 891 added across 1156 non-merge commits, @objectstack/console: minor, 98 declared-breaking entries listed
sdui.manifest.json + scripts/sdui-manifest.record.json	node scripts/gen-sdui-manifest-node.mjs. The manifest is BYTE-IDENTICAL (sha256 49211fee7792…, 57 components): the pinned commit declares the same published @object-ui 17.6.0 the old pin did, so only the provenance record moves. check:sdui-manifest exit 0
packages/sdui-parser/objectui-lockstep.json	pnpm gen:sdui-lockstep against a worktree parked at the new pin (--update refuses off the pin). See the RED below
27 .objectui-sha citations under packages/spec/src/**	each re-MEASURED against the new pin's tree — see the table
content/docs/references/ui/{dataset,view}.mdx	gen:docs, from the two corrected describes
.changeset/17429-pin-bump-describe-corrections.md	@objectstack/spec: patch for those two sentences

Lockfile / override state: pnpm install in this worktree produced no pnpm-lock.yaml change, and the bump script writes none — the console SPA is vendored as a built dist/, not as a dependency, so this bump moves no dependency edge in this repo.

Fold-in: the two-code sdui-parser port (ruling batch #201 item 1, maintainer 「201 A」)

Commit 6ef8344645 brings PR #18723's port onto this branch verbatim. Eight files, each byte-identical to the blob on claude/issue-17645-sdui-lockstep-port (git hash-object here vs git rev-parse claude/issue-17645-sdui-lockstep-port:PATH there — all eight SAME, no hunk rewritten):

path	blob
packages/sdui-parser/src/kanban-quick-add.ts	20f9f74b9f05
packages/sdui-parser/src/validate.ts	156a264073f7
packages/sdui-parser/src/types.ts	567bc1259bea
packages/sdui-parser/src/index.ts	2ea35795b9cf
packages/sdui-parser/src/codegen.ts	76a7d654f1fd
packages/sdui-parser/src/__tests__/inert-quick-add.test.ts	723cad21333a
packages/sdui-parser/src/__tests__/member-type-mismatch.test.ts	4b87cadaa067
.changeset/17645-sdui-parser-lockstep-port.md	7c171fddfc43

Deliberately NOT brought from that branch: .objectui-sha, scripts/sdui-manifest.record.json (both already this branch's, at 87af769e9a3e) and packages/sdui-parser/objectui-lockstep.json.

The lockstep record was neither text-merged nor re-recorded — and needed neither

This branch already carries the record taken at the live pin in the pin-bump round: objectui.rev = recordedAgainstPin = 87af769e9a3ee28ace099fdd653d3ebd79fe82e2, 26 diagnostic codes including inert-quick-add and member-type-mismatch. --update re-reads objectui's side only, so a re-record here would have rewritten the same bytes. Attempted anyway, and it REFUSED, which is the generator working: pnpm gen:sdui-lockstep exit 1, [head-off-pin] the objectui checkout is at c2f0f48329de, and .objectui-sha names 87af769e9a3e; record file byte-unchanged across the attempt (e0a6988b1c9a before and after). The remedy it prints — parking the sibling checkout on the pin — was not taken: that is a git write inside a read-only checkout this container shares with other agents, and the record it would produce is the one already committed.

Result — check:sdui-lockstep exit 0 on the head: OK — this copy is byte-identical to objectui@87af769e9a3e over 214 grammar line(s) [blob 0131f27cf86d] and agrees on all 26 diagnostic code(s), across 8 non-test source(s). --self-test exit 0 alongside. check:sdui-manifest exit 0 (manifest intact, sha256 49211fee7792, recorded at the live pin).

The type-check red PR #18723 carried is closed by the pin, not by a code change

The red was check:objectui-pin-citations inside the Type Check · source gates job. It reds exactly on the two heads of #18723 that claimed .objectui-sha = cfcc17d9dd04; on this branch, with the live pin, the gate is exit 0 (27 asserting citations match, 48 historical recorded and not checked, 7 anchor content assertions verified). Reverse-verified here through scripts/ablation-replace.mjs: rewriting .objectui-sha to cfcc17d9dd04 turns the gate RED with five anchor-content mismatches; restored and proven (blob back to 035b6937d641, git diff HEAD empty). So the fix is the fold direction itself: the port lands on the branch whose citations are already re-measured at the pin. No edit was made inside packages/sdui-parser/** for it, and no @ts-ignore or any was added anywhere; tsc --listFiles confirms both new test files are inside the package program.

Ported from cfcc17d9, re-verified against the LIVE pin

objectui DID move packages/sdui-parser between cfcc17d9dd04 and 87af769e9a3e: one unrelated feature, objectui#6771's retired body child-list dialect (body-dialect.ts +99, index.ts +1, validate.ts +31/−7, its own test). It touches neither ported code and stamps only not-a-container and unknown-prop, both already in the 26-code set — which is why 26 = 26 holds at the live pin with it unported. Against objectui at 87af769e9a3e the whole non-comment divergence of the five ported sources is four items, three pre-existing (ValidationResult vs ManifestValidationResult; the code: 'inert-quick-add' inline literal forced by check:dispatcher-error-vocabulary, pinned equal to the constant by a test row; provenance.ts and assertFullyLoaded / RegistryConfigLike.lazy absent here) and one new since cfcc17d9 (body-dialect.ts absent here — see Acceptance notes). The port strictly REDUCES divergence in every file it touches: validate.ts 146 → 73, index.ts 114 → 92, codegen.ts 30 → 3, types.ts 96 → 85 diff lines vs objectui at the live pin.

Ablations — re-run on this tree

Three, each through scripts/ablation-replace.mjs (anchor ×1, mutation proven on disk by anchor count and blob change, restore proven by blob == HEAD AND an empty git diff HEAD): (1) checkKanbanQuickAdd stamps nothing — 5 failed / 8 passed of 13; (2) the validator's member check reads no declared arms — 7 failed / 14 passed of 21; (3) the codegen narrows no member type — 3 failed / 18 passed of 21. Baselines 13 and 21 passed; in each leg the failures are the SUBJECT rows and every control row stays green. Both suites import ../index.js, a relative intra-package specifier, so no rebuild is in the resolution path.

Citation spelling fix (fc25a5925d)

CI's Lint & Repo Gates went red on check-issue-citations at 6ef8344645: two issue-number citations on docblock lines this branch rewrote. packages/spec/src/ui/view.zod.ts — objectstack#17328 respelled bare #17328 (the issue exists, closed completed by merged PR #17671; the gate probes only unqualified citations, so the this-repo-qualified spelling was judged against a board never asked for it). packages/spec/src/ui/component.zod.ts — #10274 measured deleted by --probe-cause (gone from the board and 404 on the web endpoint, not transferred); no replacement guessed — the sentence now states in prose that the number no longer resolves and names the live record (that block plus check:objectui-pin-citations). node scripts/check-issue-citations.mjs exit 0 at the head (31 judged, 0 findings); check:generated unchanged (TSDoc, not .describe()).

Post-landing smoke evidence (triage's condition, issue comment 5690507317)

bash scripts/pm/os-verify-lock.sh -c 'bash scripts/build-console.sh'
  os-verify-lock: VERDICT command-exit 0 · held the lock 653s (10m53s) · waited 0s
  ✓ 8786 modules transformed.
  ✓ Bundle canary 'import/jobs' present — framework client is in the bundle.
  ✓ Console bundle carries THIS tree's @objectstack/spec, and only it.
  ✓ @objectstack/console dist ready (60228 KB) from objectui@87af769e9a3e

Reading	Value
exit code	0
packages/console/dist size	60228 KB — 51,702,779 bytes, 3,285 files
content hash	sha256 259d2dfecaa12507fd3166a3f91c66c52581990e5ccac3899757027fa6ac108d over find … -type f | sort | xargs sha256sum | sha256sum
provenance stamp	packages/console/dist/.objectui-sha = 87af769e9a3ee28ace099fdd653d3ebd79fe82e2
pnpm check:console-sha	exit 0 — "Console dist matches the objectui pin"

The console build is GREEN at the new pin. 584 commits of objectui change came with it and none of them broke it.

The citations — re-MEASURED, never re-pointed

27 asserting citations, all of them stale at the new pin. Each was re-derived by reading objectui at BOTH shas and comparing content, not by arithmetic on a line number. pnpm --filter @objectstack/spec check:objectui-pin-citations exit 0; --verify-anchors exit 0 with 7 of 7 content assertions verified against objectui at 87af769e9.

Corpus-absence records (7 citations + 6 generated copies)

These cite a zero-hit count over the pinned tree, not a file:line. Re-counted with git grep -F -o at both shas; the METHOD was calibrated first by reproducing the old pin's own numbers exactly (6409 tracked files, useState 2304).

Record	Cited	Reads at 87af769e9	Verdict
18.kernel-health-check-and-hot-reload-durations-unit-in-key.ts:48	13 exports of plugin-lifecycle-advanced.zod.ts + debounceDelay = 0 / 6409 files; controls objectstack 10171, @objectstack/spec 3479	all still 0 / 8228 files; controls 12966, 4997	unchanged verdict, counts re-measured
18.kernel-runtime-config-timeout-unit-in-key.ts:45	resourceLimits.timeout 0; controls timeout 832, RuntimeConfig 236, resourceLimits 2	0; 1075, 240, 2 — and both resourceLimits hits are still prose in packages/app-shell	unchanged
18.logging-durations-unit-in-key.ts:48	4 names 0; controls useState 2304, timeout 702	0; 2383, 1075	unchanged — ⚠️ but timeout 702 was WRONG AT THE PIN IT NAMED: that corpus reads 832. An error since written, found only by re-measuring, corrected here
18.system-metrics-jsdoc-durations-unit-in-key.ts:70	6 def names + 2 keys 0; controls window 2710, timeout 832, period 160, interval 156, metrics 301	all 0; 3464, 1075, 170, 170, 324	unchanged
18.system-tracing-otel-exporter-durations-unit-in-key.ts:66	37 exports of tracing.zod.ts + 4 key names 0; Span 404 / SpanSchema 40 unrelated; controls 10171, 3479	all 0; 486 / 53; 12966, 4997	unchanged
18.tenant-schema-cache-ttl-unit-in-key.ts:31	schemaCacheTTL 0; controls TTL 112, tenant 819	0; 156, 976	unchanged
migrations/registry.ts ×6	the generated concatenation of the six above	regenerated with pnpm --filter @objectstack/spec gen:migration-registry	follows its entries

Read-point records (20 citations + 7 content assertions)

Record	Anchor, as cited	At 87af769e9	Verdict
data/api-methods-batch-conformance.test.ts:64	ObjectGrid.tsx:3538-3553	:3940-3955	MOVED, span byte-identical (git hash-object 6133933199… both sides)
	hooks/useBulkExecutor.ts:284-289	:298-303	MOVED, span byte-identical (0108334833…); ⚠️ the FILE is no longer byte-identical (+36/-22), so the record's old identity argument is gone and the span was re-read
ui/component.zod.ts tab-item icon + ui/component.test.ts:376	containers.tsx:730-736	:853-859	MOVED, 7 lines byte-identical
	containers.tsx:789 (registration items input)	:912	MOVED and the LINE CHANGED — it lost label: 'Tabs', gained of: 'object' and a longer description; the member list the record cites is unchanged
ui/component.zod.ts accordion-item icon + component.test.ts:286	containers.tsx:919-925	:1069-1075	MOVED, byte-identical
	containers.tsx:966	:1116	MOVED and the LINE CHANGED, same way
ui/component.zod.ts button icon + component.test.ts:3489	form/button.tsx:43, :72, :74	unchanged	UNCHANGED
	form/button.tsx:85-102 (inputs), :103-107 (defaultProps)	:85-97, :98-102	MOVED and SHRANK — every input dropped its label and defaultValue; the four input names and the ABSENCE of an icon input, which is what the record asserts, both hold
	action/resolve-icon.ts:129-132 (resolveIcon)	:322-328	MOVED and REWRITTEN — the tail no longer indexes lucide-react's icons record; it asks recordIconName and hands the pair to lazyIconComponent. Accept/reject behaviour unchanged
	resolve-icon.ts:117-120, :100-105, :90-92	:302-305, :153-158, :143-145	MOVED, byte-identical
	lib/lazy-icon.tsx:66-92	:98-124	MOVED, byte-identical
ui/component.zod.ts object-metric icon	plugin-dashboard/src/index.tsx:204	:237 (registration now opens at :227)	MOVED and the LINE CHANGED — the input's label: 'Icon (Lucide name)' is GONE; the key is still published, now as a bare { name: 'icon', type: 'string' }
	ObjectMetricWidget.tsx:142 / :474; MetricWidget.tsx:312-321 / :373-382; lazy-icon.tsx:66-80	:174 / :483; :351-360 / :412-421; :98-112	MOVED, byte-identical
ui/component.zod.ts kanban limit ×2 + component.test.ts:3360	ObjectKanban.tsx:264 ($top: schema.limit ?? DEFAULT_KANBAN_LIMIT)	:676 ($top: resolveRowLimit(schema.limit, DEFAULT_KANBAN_LIMIT))	MOVED and REWRITTEN — objectui#9925 put a refusal in front of it (a contract-refused cap is dropped and reported at :553). The pinned fact, that limit lowers into the query's top-level $top, holds
	ObjectKanban.tsx:71; plugin-kanban/src/index.tsx:395-398	:84; :447-450	MOVED, byte-identical
	ObjectKanban.tsx:143 + plugin-kanban/src/types.ts:134 (KanbanSchema.limit?: number)	⚠️ VANISHED	KanbanSchema was RETIRED at this pin (maintainer ruling 2026-09-09) and types.ts declares the member no more. Re-derived rather than re-pointed: the published twin is ObjectKanbanSchema, imported at ObjectKanban.tsx:10, declaring limit?: number at packages/types/src/objectql.ts:3735
	ElementDataSourceGate.tsx:236-241	:316-331	MOVED and EXTENDED — the branch gained objectui#9899's presence-is-not-authorship test and a describeDisplacedRowLimit report
ui/component.zod.ts kanban quickAdd + component.test.ts:3427	ObjectKanban.tsx:931; plugin-kanban/src/index.tsx:196; KanbanImpl.tsx:355 / :368	:1563; :313; :621 / :634	MOVED, byte-identical. The absence re-counted at this pin: quickAdd / onQuickAdd 0 each in ObjectKanban.tsx, against 11 (was 6) for the sibling onCardClick
ui/dataset.zod.ts:207 measure format	dataset-format.ts:185-198, date-display.ts:131-164 / :117, :195	:229-264 (routed at :370), :198-233 / :152, :260-262	⚠️ SUBSTANCE REVERSED — see below
ui/view.zod.ts:2945 FormField.span	the widest-tier-only class claim at 53ded82bf7	spanLadderFor, form/form.tsx:204-231	⚠️ SUBSTANCE REVERSED — see below
ui/view.zod.ts ListMapConfig, 7 content assertions	ListView.tsx:113 / :67; ObjectView.tsx:1381; objectql.zod.ts:562 / :313; ObjectMap.tsx:373 / :378	:146 / :85; :1764; :1574 / :734; :385 / :390	all 7 MOVED; ⚠️ one LOST ITS SYMBOL — FLAT_MAP_CONFIG_KEYS became the total FLAT_MAP_CONFIG_SPELLING map (objectui#9950), so style now DOES reach the flat product, as mapStyle. The quote was re-read, not re-pointed

The two records whose SUBSTANCE the range reverses

Both are claims about what the SHIPPED renderer does, so the pin is what dates them, and both were re-stated from the new tree. Each also carried a published .describe() saying the same now-false thing, which this bump is what falsifies — so both sentences are corrected here and content/docs/references/ui/{dataset,view}.mdx regenerated from them. That is the @objectstack/spec: patch changeset in this PR.

• Dataset measure format said "a datetime value ignores it". objectui#8352 is inside the range: formatMeasureDate's datetime arm now SELECTS a formatter — relative to formatRelativeDate (:260), short to formatDateTime(v, { locale, style: 'compact' }) (:261), everything else to the bare formatDateTime(v, { locale }) (:262). A date PATTERN is still ignored on both arms, which is the half that survives.
• FormField.span said only the widest container-query tier's class is emitted, so a 'full' field was one cell of two at 720px ([finding] absolute-colspan-discouraged asserts a misalignment that does not happen — measured in Chromium at all three surface widths, and its recommended alternative span: 'full' is the spelling that misrenders #17328). objectui#9244 / objectui#9253 (bd09957380) are inside the range: spanLadderFor emits one clamped col-span class per multi-column tier. ⭐ The previous revision of this block asked in writing to be re-read at the bump that absorbed it; this is that re-read. The 'auto' half — textarea, markdown, html, richtext and repeater — re-measured UNCHANGED.

Gates

Derived with node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstack (122 commands for this 19-path, 757-line change set). Run locally, exit codes captured before any pipe:

pnpm check:sdui-lockstep                                        EXIT 0   (after the fold-in: 26 = 26 codes, grammar region byte-identical)
pnpm --filter @objectstack/spec check:objectui-pin-citations    EXIT 0   (27 asserting citations match, 7/7 anchors verified)
  … --verify-anchors                                            EXIT 0
node scripts/check-sdui-manifest.mjs                            EXIT 0
pnpm check:objectui-bump                                        EXIT 0
pnpm check:objectui-changeset                                   EXIT 0
pnpm check:console-sha                                          EXIT 0
pnpm --filter @objectstack/spec check:generated                 EXIT 0   (15/15 artefacts current)
MANIFEST=… check:react-declaration-parity                       EXIT 0   (reading unmoved: the manifest is byte-identical)
node scripts/check-adr-0087-registration.mjs --base origin/main EXIT 0
node scripts/check-changeset-no-major.mjs --base origin/main    EXIT 0
node scripts/check-empty-changeset.mjs --base origin/main       EXIT 0
pnpm lint                                                       EXIT 0   (repo-wide, eslint . --no-inline-config)
pnpm --filter @objectstack/spec test                            EXIT 0   (503 files, 14710 tests)
pnpm --filter @objectstack/spec --filter @objectstack/sdui-parser typecheck   EXIT 0

plus 27 further derived families, every one exit 0: check:nul-bytes, check:merge-driver, check:watch-hint-literal, check:comment-mask-adoption, check:comment-mask-corpus, check:doc-frontmatter, check:docs-section-name, check:scripts-symbol-anchors, check:spec-docblock-symbol-anchors, check:ci-filter-parity, check:keyed-text-bounds, check:undeclared-dep-imports, check:section-landing-index, check:liveness, check:duration-unit-keys, check:llms-txt, check:spec-changes, check:upgrade-guide, check:strictness-ledger, check:export-origins, check:api-surface, check:authorable-surface, check:docs, check:doc-anchors, check:docs-spec-enumerations, check:quick-reference-counts, check:corpus-claim-drift, check:changeset-gate-self-tests, check:pm-widening-tells, check:published-files, and both changeset self-tests.

The rest of the 122 are left to CI, which runs the whole farm. origin/main moved 7 commits while this was built and touches NONE of this diff's 19 paths, so no merge was taken; the queue's rebuilt generation is what validates the join.

Acceptance notes

• check-issue-citations blind spot (measured on [finding] absolute-colspan-discouraged asserts a misalignment that does not happen — measured in Chromium at all three surface widths, and its recommended alternative span: 'full' is the spelling that misrenders #17328 at 6ef8344645): a citation qualified with this repository's own name (objectstack#N or owner/repo#N) is recognised as naming THIS repo and resolved against the board, but the board is only asked for unqualified citations and --probe-cause's guard is !cite.qualifier too — a live issue reads allocated-but-absent and NOT MEASURED under --probe-cause. Not fixed here (the gate is not this PR's surface); recorded for the next author of scripts/check-issue-citations.mjs.

• objectui's packages/sdui-parser/body-dialect.ts (objectui#6771, landed between cfcc17d9dd04 and the live pin) has no counterpart in this copy: for an authored body key the two copies stamp different codes (not-a-container under a non-container there, the bare unknown-prop here), invisible to check:sdui-lockstep's code-SET comparison because both codes are already in the 26-set. Both severities are warning, so compile().ok and the save verdict are unchanged. A third code path the ruling does not name; carrier: the next sdui-parser port round.

• provenance.ts and assertFullyLoaded / RegistryConfigLike.lazy remain absent from this copy (zero diagnostic codes; sdui-parser: port inert-quick-add and member-type-mismatch from objectui, re-record the lockstep and bump the pin #18723's acceptance notes already record provenance.ts). pnpm gen:sdui-lockstep is unrunnable in this container because the shared sibling checkout sits off the pin and the generator's remedy is a git write in someone else's tree; it cost nothing this round and will bite the next pin bump, which must park its own checkout.

• .changeset/18516-span-auto-wide-types.md (pending, not yet released) describes the FormField.span describe text this PR corrects, including "At this pin only the widest tier's class is emitted". check:empty-changeset refuses any diff that MODIFIES a changeset present at the merge base, so it is deliberately left untouched here and reported instead.

• The element:button icon .describe() still says the renderer resolves through lucide-react's icons map. At the new pin that is one hop indirect (recordIconName + lazyIconComponent) but the glyph set and the accepted spellings are unchanged, so it was left as written rather than churned.

Original author: the director seat's os-dev rounds (pin bump; fold-in 6ef8344645; citation fix fc25a5925d), session_012GcsUbuqFGBibkEDMRC1eE; body maintained by the director seat, session_012GcsUbuqFGBibkEDMRC1eE.

---

Generated by Claude Code

[github-actions]

📓 Docs Drift Check

This PR changes 2 package(s): @objectstack/sdui-parser, @objectstack/spec, touching 22 documentable anchor(s). ⚠️ 1 changed file(s) yielded no anchor (packages/sdui-parser/objectui-lockstep.json), so the pages documenting them are NOT COVERED by this run — this is not a clean bill of health for those files.

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

• content/docs/data-modeling/analytics.mdx (via DatasetMeasureSchema (symbol, a top-level const))

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

• content/docs/releases/v15.mdx (via PageTabsProps (symbol, a top-level const object))

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

• 1 changed file(s) yielded no anchor (packages/sdui-parser/objectui-lockstep.json) — pages documenting those are invisible to this run
• 3 name(s) were too generic to anchor anything (single lowercase words)
• the SDK route bridge reached 60 of 215 client-bound route-ledger rows — the other 155 have no registrar path: tail to select them, so pages documenting THEIR client methods cannot appear above, on this or any run. Of those 155: 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; 100 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 — 137 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 50bae37f3df71e759f1730b906d5af35baa7b0f7 → packageMentionDocs.

Which tree this was computed on

This run read content/docs from ff53bb6f7f47a676c6b47bd379df9e37c706ece0 — the merge of head fc25a5925d7d6b55ed25b81d5f1fcef4be836109 into base 50bae37f3df71e759f1730b906d5af35baa7b0f7, 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 ff53bb6f7f47a676c6b47bd379df9e37c706ece0 && git checkout ff53bb6f7f47a676c6b47bd379df9e37c706ece0
# afterwards, rebuild it from the two parents, which stay fetchable
git fetch origin 50bae37f3df71e759f1730b906d5af35baa7b0f7 fc25a5925d7d6b55ed25b81d5f1fcef4be836109 && git checkout -B drift-repro 50bae37f3df71e759f1730b906d5af35baa7b0f7 && git merge --no-ff fc25a5925d7d6b55ed25b81d5f1fcef4be836109

node scripts/docs-audit/affected-docs.mjs --json 50bae37f3df71e759f1730b906d5af35baa7b0f7

⚠️ 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 50bae37f3df71e759f1730b906d5af35baa7b0f7 → pass the list as
args.docs, on the commit named under Which tree this was computed on.

[os-project-manager]

Director seat (session_012GcsUbuqFGBibkEDMRC1eE), 2026-09-20T16:15Z — on the red Single-Claim Path Guard (No other open PR may claim the same single-writer path, job 106107682399, head 1a5b8f3547).

What is failing: .objectui-sha is a declared single-writer path and a second open PR modifies it — PR #18723 (draft, 53ded82b… → cfcc17d9…), which the gate counts by design (「a draft claimant still counts」). Not this PR's defect: under ruling batch #200 item 4 (5750928995 on #17645) the pin lands here first and #18723 takes main's pin afterwards, so #18723 is the one that must stop claiming the path. Fix in flight: an os-dev round is restoring .objectui-sha and scripts/sdui-manifest.record.json on #18723's branch to origin/main's blobs; it will comment here with the commit, after which this gate is re-run once. #18723 stays draft, red on its own lockstep gate by design until its post-#19398 merge lap.

---

Generated by Claude Code

[os-project-manager]

Director seat (session_012GcsUbuqFGBibkEDMRC1eE), 2026-09-20T16:21Z — on the red Lint & Repo Gates (job 106107682913, head 1a5b8f3547; failing step: sdui-parser stays in lockstep with objectui's copy).

What is failing: check:sdui-lockstep reads [code-drift] — objectui at the new pin 87af769e9a3e carries 26 sdui-parser diagnostic codes, this repo 24, and the two missing are exactly inert-quick-add and member-type-mismatch, the port PR #18723 carries (the dev's report on #17429 measured it: the 24 are a strict subset, grammar region byte-identical). This PR cannot be green on its own at any pin this card may take, and #18723 cannot be green without this pin: the gate is one invariant across two PRs. ⛔ Not a flake, ⛔ not re-run. Fix pending a decision: batch #201 is in front of the maintainer — recommended letter A folds #18723's reviewed port (with its tests and ablation) into this PR, re-records the lockstep, re-takes the contract review here (Clause-② becomes yes) and closes #18723 as superseded. The push follows the letter; this PR stays draft until then.

---

Generated by Claude Code

[os-project-manager]

Director seat (session_012GcsUbuqFGBibkEDMRC1eE), 2026-09-20T16:24Z: the .objectui-sha collision is cleared on #18723's side as of commit 8a778067eb58d9b79241f6546016ea411c1269c8 — that PR no longer modifies the declared single-writer path (its file list is 9 paths, with neither .objectui-sha nor scripts/sdui-manifest.record.json among them; both restored byte-identical to main's blobs, check-sdui-manifest exit 0 there). The Single-Claim Path Guard re-judges this PR on its next event; the Lint & Repo Gates red (lockstep [code-drift]) stays as recorded above until batch #201 is answered.

---

Generated by Claude Code

[os-project-manager]

Director seat (session_012GcsUbuqFGBibkEDMRC1eE), 2026-09-20T22:56Z — on the red Check Changeset (job 106164107369, step 「Guard against accidental major bumps (launch window)」, head 1a5b8f3547, run triggered by the needs:contract-review label this seat hung at 2026-09-20T22:54Z).

What is failing: the guard read the carrier (needs:contract-review on the PR) against the body line as it was then (Clause-②: no) and, taking the carrier as the declaration, asked for the widened package to be graded minor or above; the only package whose packages/**/src/** this head moves is @objectstack/spec, graded patch in .changeset/17429-pin-bump-describe-corrections.md (two .describe() sentences — correctly patch). Why it turns green without touching that grade: under ruling batch #201 A (5753263180 on #17645) the fold-in commit now in flight adds #17645's sdui-parser port with its own changeset .changeset/17645-sdui-parser-lockstep-port.md (@objectstack/sdui-parser: minor) and moves packages/sdui-parser/src/** — the package that actually grows — which is the 「one minor on a package this diff moved」 the gate names as clearing this red. The body's declaration line already reads Clause-②: yes (rewritten by this seat). ⛔ Not re-run before that push; the push re-runs it.

---

Generated by Claude Code

[os-project-manager]

Contract review

Served-tier: CONTRACT_REVIEW_TIER
Head-sha: fc25a5925d7d6b55ed25b81d5f1fcef4be836109

Isolated at-tier reviewer; reviewed 2026-09-21T00:13Z. Merge-base a88a97332fc90cb04e9dbae2d4e2f5aa56cab708; origin/main tip during the review 50bae37f3df71e759f1730b906d5af35baa7b0f7 (the PR payload's base sha c5d3d1d8b6 is older; no merge was judged); worktree ../objectstack-review-19398 detached at the head, pnpm install exit 0, removed at the end; objectui read READ-ONLY from the sibling checkout at 87af769e9a3ee28ace099fdd653d3ebd79fe82e2 (new pin), 53ded82bf7a494f54e344e19099dbf00854b8694 (old pin) and cfcc17d9dd04 (the revision PR #18723 ported from), via git show / git grep / git archive into scratch; the console build's own git worktree add of objectui at the pin under the review worktree's .cache/ is the script's documented mechanism and was removed with the worktree. Every exit code below was captured before any pipe; the three os-dev-report comments were treated as claims and re-derived, never adopted.

① Derived judgments

1. The pin — RIGHT. .objectui-sha at the head reads 87af769e9a3ee28ace099fdd653d3ebd79fe82e2 (blob 035b6937d641). git -C ../objectui merge-base --is-ancestor cfcc17d9dd04 87af769e9a3e exit 0 (and 53ded82bf7a4 is an ancestor too; 1156 non-merge commits in the range, the number the console digest names). pnpm check:objectui-bump exit 0 (20 assertions / 5 cases); pnpm check:objectui-changeset exit 0; pnpm check:console-sha exit 0; node scripts/check-sdui-manifest.mjs exit 0 and --self-test exit 0 (11 cases). Manifest re-hashed by me: sha256sum sdui.manifest.json = 49211fee7792cf51174930dc2c1be2169d5175f77a83491f68f90b3f0c19e69d, 57 components, byte-identical to origin/main's manifest; scripts/sdui-manifest.record.json moves only objectuiSha and generatedAt, and @object-ui packages/core/package.json reads 17.6.0 at BOTH shas, so the byte-identical manifest is expected, not suspicious.

2. The citations — RIGHT. pnpm --filter @objectstack/spec check:objectui-pin-citations exit 0 (27 asserting match 87af769e9, 48 historical, 1497 spec sources, 7 anchor content assertions verified); --verify-anchors exit 0. I re-derived 21 of the 27 rows myself by reading objectui at BOTH shas and comparing content, plus every one of the 7 content assertions:

   • Corpus-absence, all six records: the calibration reproduces the OLD pin's own numbers exactly with git ls-tree -r and git grep -F -o (6409 tracked files; useState 2304; objectstack 10171; @objectstack/spec 3479; RuntimeConfig 236; resourceLimits 2; TTL 112; tenant 819; window 2710; period 160; interval 156; metrics 301; Span 404; SpanSchema 40) and timeout reads 832 there, so the logging record's old control 702 was indeed WRONG at the pin it named. At 87af769e9 every control matches the PR's table (8228 files; 2383; 12966; 4997; 1075; 240; 2, both hits prose in packages/app-shell; 156; 976; 3464; 170; 170; 324; 486; 53) and every zero-hit token is still 0 (schemaCacheTTL, debounceDelay, resourceLimits.timeout, flushInterval, initialDelay, HttpDestinationConfig, LoggingConfig, and the seven distinctive metrics keys). Six rows RIGHT; the registry.ts ×6 rows follow them.
   • api-methods-batch-conformance.test.ts: ObjectGrid.tsx OLD 3538-3553 and NEW 3940-3955 both hash 6133933199230670e29d8c7f51c558d86a0af1d2 (file diffstat 721/159 as stated); useBulkExecutor.ts OLD 284-289 and NEW 298-303 both hash 01083348330f10a201cdf1078b4c21c236402b6a, file 36/22, :303 still label = 'bulk delete'. RIGHT ×2.
   • tab-item icon: layout/containers.tsx OLD 730-736 and NEW 853-859 hash c542d07a70bf both; registration OLD :789 carried label: 'Tabs', NEW :912 carries of: 'object' and no label, member list unchanged. Accordion-item: 919-925 and 1069-1075 hash 3934c0ae547e both; :966 to :1116 the same change. RIGHT ×2.
   • element:button icon: form/button.tsx :43, :72, :74 byte-unchanged; inputs OLD 85-102 and NEW 85-97 both name exactly label, variant, size, className (no icon input), label:/defaultValue: members 6 → 0; defaultProps NEW 98-102. resolve-icon.ts OLD 129-132 indexes icons[…], NEW 322-328 asks recordIconName then lazyIconComponent (0 icons[ in the tail); the three moved spans (117-120→302-305, 100-105→153-158, 90-92→143-145) and lazy-icon.tsx 66-92→98-124 hash identical. RIGHT.
   • object-metric icon: plugin-dashboard/src/index.tsx OLD :204 { name: 'icon', type: 'string', label: 'Icon (Lucide name)' }, NEW :237 { name: 'icon', type: 'string' }, registration opens at :227. RIGHT.
   • kanban limit: ObjectKanban.tsx OLD :264 $top: schema.limit ?? DEFAULT_KANBAN_LIMIT, NEW :676 $top: resolveRowLimit(schema.limit, DEFAULT_KANBAN_LIMIT) with the refusal message at :553; :71→:84 DEFAULT_KANBAN_LIMIT = 100; plugin-kanban/src/index.tsx 395-398→447-450 hash identical; plugin-kanban/src/types.ts at the pin declares NO KanbanSchema interface (its :35 is the by-name tombstone) and ObjectKanban.tsx:10 imports ObjectKanbanSchema from @object-ui/types, whose packages/types/src/objectql.ts:3735 reads limit?: number. ElementDataSourceGate.tsx OLD 236-241 → NEW 316-331 gained isUsableRowLimit and describeDisplacedRowLimit. RIGHT.
   • kanban quickAdd: plugin-kanban/src/index.tsx :196→:313 and KanbanImpl.tsx :355/:368→:621/:634 byte-identical; ObjectKanban.tsx :931→:1563 is the same ...effectiveSchema, spread (only JSX indentation moved two columns, the cited text is identical); quickAdd 0, onQuickAdd 0, onCardClick 11 (was 6) in ObjectKanban.tsx at the pin. RIGHT.
   • ListMapConfig, all 7 content assertions: ListView.tsx:146 function resolveListMapConfig…, :85 export const FLAT_MAP_CONFIG_SPELLING = { (OLD :67 FLAT_MAP_CONFIG_KEYS listed seven keys satisfies … Omit ObjectMapConfig, 'style'; NEW maps all eight with style: 'mapStyle'), ObjectView.tsx:1764 case 'map':, objectql.zod.ts:1574 ObjectMapConfigSchema = z.object({ declaring the eight keys and :734 LIST_VIEW_LOCAL_OVERRIDES, ObjectMap.tsx:377 / :385 / :390, plugin-map.mdx:143 — each line reads exactly as quoted, and each old anchor reads the same at the old pin. RIGHT ×7.
   • Dataset measure format (SUBSTANCE REVERSED) — RIGHT, checked against the code, not the prose: at 87af769e9 packages/core/src/utils/dataset-format.ts formatMeasureDate spans :229-264 (routed at :370) and its datetime arm now reads if (format === 'relative') return formatRelativeDate(v, { locale }); (:260), if (format === 'short') return formatDateTime(v, { locale, style: 'compact' }); (:261), else bare formatDateTime(v, { locale }) (:262); at 53ded82bf7 the same function (:185-198) returned formatDateTime(v, { locale }) unconditionally (:195). date-display.ts formatDate at :198 (was :131) and the ±7-day fallback at :152 (was :117, now through absoluteFallbackOptions). The new .describe() ("a date or datetime value reads format as a display style — short or relative, honoured on both"; a date pattern still ignored) states exactly that code.
   • FormField.span (SUBSTANCE REVERSED) — RIGHT, checked against the code: packages/components/src/renderers/form/form.tsx declares function spanLadderFor(containerClass, targetCols) at :204 (body through :231, closing brace :232), which walks the container class's grid-cols-N tiers and pushes one clamped SPAN_CLASS[tier:min(tier.cols, targetCols)] per tier that can give more cells than already spanned; autoLayout.ts:153 is export function resolveColSpan(. The old pin's form.tsx (:2482-2495) picked a single class. The new .describe() ("one clamped col-span class per multi-column tier … the whole row at every multi-column tier, not just the widest") states that code; the 'auto' half is unchanged.

3. The fold-in — RIGHT. All eight files hash SAME between the head and origin/claude/issue-17645-sdui-lockstep-port (kanban-quick-add.ts 20f9f74b9f05, validate.ts 156a264073f7, types.ts 567bc1259bea, index.ts 2ea35795b9cf, codegen.ts 76a7d654f1fd, inert-quick-add.test.ts 723cad21333a, member-type-mismatch.test.ts 4b87cadaa067, the changeset 7c171fddfc43); the port branch's only other file, objectui-lockstep.json, was correctly not brought. node scripts/check-sdui-lockstep.mjs exit 0 and --self-test exit 0. Re-derived with the head's own fingerprintOf against git archive extracts of objectui: THIS COPY 26 codes / 8 files; objectui@87af769e9a3e 26 codes / 10 files (the two extra are body-dialect.ts and provenance.ts); objectui@cfcc17d9dd04 26 / 9. The 26 on both sides, identical: bad-attr bad-attr-value bad-self-close bad-tag expected-element expression-child forbidden-attr forbidden-tag inert-expression inert-quick-add invalid-enum member-type-mismatch mismatched-tag missing-required-prop multiple-roots no-root not-a-container type-mismatch unclosed-comment unclosed-element unconsumed-widget-option unknown-component unknown-prop unterminated-brace unterminated-open-tag unterminated-string; grammar region 214 lines, blob 0131f27cf86d on both sides. pnpm --filter @objectstack/sdui-parser test through the lock: VERDICT command-exit 0, 176 passed / 10 files; typecheck VERDICT command-exit 0. Ablations re-run here through scripts/ablation-replace.mjs (anchor ×1, mutation proven by anchor 1 → 0 and blob change, restore proven blob == HEAD and empty git diff HEAD): baselines 13 passed and 21 passed; (1) checkKanbanQuickAdd returns null unconditionally (20f9f74b9f05 → befe6d71594c) — 5 failed / 8 passed (13); (2) inputTypeArms(undefined) in the validator's member check (156a264073f7 → 6a6f79d7736c) — 7 failed / 14 passed (21); (3) inputTypeArms(undefined) in memberTsType (76a7d654f1fd → 2dcbc5b677bd) — 3 failed / 18 passed (21); all three blobs back at HEAD and git status clean. @ts-ignore / @ts-expect-error / : any / as any additions over the code files of the diff: 0 (the three as any matches in the raw diff are prose lines of the console changeset digest).

4. The citation-spelling fix — RIGHT. git diff 6ef8344645 fc25a5925d --stat is exactly packages/spec/src/ui/component.zod.ts (+9/−1) and packages/spec/src/ui/view.zod.ts (+1/−1), 2 files, +10/−2. node scripts/check-issue-citations.mjs at the head exit 0 (31 judged / 15 files, 30 cross-repo-unjudged, 1 resolves, 0 findings); the proxy hint appeared and --use-env-proxy exit 0 with the same reading; --probe-cause --use-env-proxy exit 0 (nothing left to probe on the head's added lines, by design). My own probe of [finding] The #9881/#9972 read-point records still cite .objectui-sha = 82a94170c — #10137 moved the pin, so "the pin this repo builds against" is now false in spec prose and tests #10274: GET /repos/objectstack-ai/objectstack/issues/10274 HTTP 404 with no redirect, and the web endpoint HEAD 404 with no redirect — a transferred issue keeps a web redirect, so DELETED is the accurate class and the tombstone prose in component.zod.ts:2135-2143 ("minted, gone from the board and 404 on the web endpoint too — DELETED, not transferred", no replacement guessed, the live record named) states it accurately. GET /issues/17328 HTTP 200: closed, completed, an issue not a PR, so the bare #17328 is a resolving citation.

5. Merge faithfulness — RIGHT. git diff origin/main...fc25a5925d --name-only is 27 paths and equals the PR's file list as a set (0 only-PR, 0 only-diff). node scripts/check-empty-changeset.mjs --base origin/main exit 0 (no merge-base changeset modified or deleted). pnpm --filter @objectstack/spec build && pnpm --filter @objectstack/spec check:generated through the lock: VERDICT command-exit 0, held 197s; "Checking 15 generated artifacts (every gate runs)" then "✓ All 15 generated artifacts are up to date", with check:react-declaration-parity reported EXTERNAL_INPUT_REQUIRED by that aggregate as designed; git status --porcelain --untracked-files=no reads 0 tracked changes afterwards, so nothing was regenerated by the check and the committed content/docs/references/ui/{dataset,view}.mdx are the ones gen:docs produces from the two corrected sentences.

6. The console — RIGHT. bash scripts/pm/os-verify-lock.sh -c 'bash scripts/build-console.sh' in the worktree: os-verify-lock: VERDICT command-exit 0 · held the lock 642s (10m42s) · waited 0s; ✓ 8786 modules transformed.; ✓ Bundle canary 'import/jobs' present; ✓ Console bundle carries THIS tree's @objectstack/spec, and only it.; ✓ @objectstack/console dist ready (60240 KB) from objectui@87af769e9a3e. packages/console/dist/.objectui-sha = 87af769e9a3ee28ace099fdd653d3ebd79fe82e2; pnpm check:console-sha exit 0 ("Console dist matches the objectui pin"). My dist is 51,713,761 bytes / 3285 files against the dev's 51,702,779 / 3285 — same file count, not byte-reproducible across two builds, which no gate asserts.

7. CI on the head — RIGHT (all seven required contexts green). GET /commits/fc25a5925d…/check-runs read 2026-09-21T00:12Z, 42 runs, one page: TypeScript Type Check success (completed 2026-09-20T23:59Z), Test Core success (2026-09-21T00:02Z), Dogfood Regression Gate success (2026-09-20T23:53Z), Build Core success (2026-09-20T23:51Z), Temporal Conformance (live PG + MySQL) success (2026-09-20T23:53Z), Lint & Repo Gates success (2026-09-21T00:07Z; it was still in progress at my two earlier reads), Governed Surface Queue Guard success (2026-09-20T23:43Z). Check Changeset success (2026-09-20T23:51Z), not red. Console Pin Gate, Type Check · source gates, Spec property liveness, both single-writer / same-issue guards and "The card this PR closes must claim this branch" all success; no run on the head is red.

② Semver level

• .changeset/console-87af769e9a3e.md @objectstack/console: minor — RIGHT by the repo's own rule: scripts/objectui-changeset-digest.mjs takes the highest level objectui declared across the 584 releasing changesets (LEVEL_RANK), and objectui's release window bans major, so the 98 BREAKING-annotated entries ride minor by construction; the console is a vendored SPA, not a spec surface, and node scripts/check-adr-0087-registration.mjs --base origin/main exit 0 accepts its not-required (no-migration-prescription) marker (no ObjectStack-authorable key moves).
• .changeset/17645-sdui-parser-lockstep-port.md @objectstack/sdui-parser: minor — RIGHT: two new diagnostic codes, a new optional ManifestInput.of member and a codegen narrowing that fires only when of is declared are additive; nothing is removed or renamed; both codes are warning, so compile().ok is unchanged. Clause-②: yes is the correct declaration for a published save-gate vocabulary widening, and the rule "yes takes at least minor" is met by this changeset; minor, not major, is right because no accept set narrows.
• .changeset/17429-pin-bump-describe-corrections.md @objectstack/spec: patch — RIGHT: the two .describe() sentences change documented renderer behaviour but no key, enum member, default or export moves (check:authorable-surface and check:api-surface current inside check:generated); a corrected sentence is a fix, and its own Clause-②: no line is accurate for the spec package's diff — the PR-level yes comes from the sdui-parser half.
• Gates: node scripts/check-changeset-no-major.mjs --base origin/main exit 0 (no major); node scripts/check-adr-0087-registration.mjs --base origin/main exit 0; pnpm check:changeset-gate-self-tests exit 0.

③ Boundary flags

• body-dialect divergence — CONFIRMED, not a blocker. objectui@87af769e9a3e packages/sdui-parser/src/body-dialect.ts (objectui#6771, commit 2acd8e109, +99 lines with index.ts +1 and validate.ts +31/−7 since cfcc17d9) makes an authored body child-list stamp not-a-container under a non-container (:82-83) and a named-replacement unknown-prop otherwise (:92-93), called from validate.ts:105 ahead of the bare unknown-prop; this copy's validate.ts:97 stamps the bare unknown-prop in both cases. Every one of those is severity: 'warning', so the save verdict and compile().ok agree — a message/code divergence inside the 26-code set, invisible to the SET comparison. Not this PR's: the ruling names two codes, and the dev's to-file class-(b) finding is the right carrier.
• provenance.ts / assertFullyLoaded absence — CONFIRMED by the fingerprintOf file lists (objectui 10 files, this copy 8; both extra files stamp zero codes). Not a blocker: no diagnostic code, and the gate compares codes, not file lists.
• check:skill-examples exit-code spelling — NOT MEASURED here (not run; nothing in this diff reaches packages/client-react). A gate-authoring note, not this PR's surface.
• gen:sdui-lockstep unrunnable off-pin — CONFIRMED structurally: the shared sibling checkout /home/user/objectui is at c2f0f48329de on a branch, not the pin; I did not run the generator (its remedy is a checkout in that shared tree). The committed record is nonetheless the one the generator would write: objectui at the pin carries exactly the 10 files and 26 codes recorded, and only rev/revDate/two files/two codes moved vs origin/main. Not a blocker.
• check-issue-citations blind spot — CONFIRMED in source: classifyCitation treats objectstack-ai/objectstack#N as this repo (:428) and falls through to the board, while the --probe-cause guard reads !cite.qualifier (:748), so a qualified this-repo citation is never probed. Not a blocker; fixed in the PR text only, as the dispatch allowed.
• .changeset/18516-span-auto-wide-types.md (pending at the merge base) carries "At this pin only the widest tier's class is emitted" — CONFIRMED (one widest hit); correctly left untouched here because check-empty-changeset refuses merge-base changeset edits.
• pnpm lint repo-wide exit 0 (run BEFORE the console build populated .cache/objectui-*, which eslint.config.mjs does not ignore). node scripts/pm/check-widening-tells.mjs --declaration yes --diff pr.diff exit 0 (a yes routes to contract review; nothing to block) and pnpm check:pm-widening-tells self-test exit 0 (508 cases).
• Carrier hygiene for the dispatching seat — node scripts/pm/check-clause2-carriers.mjs --pair 19398 exit 2: pair PR chore(objectui): bump the console pin to 87af769e9a3e with 27 re-measured citations, and fold in the two-code sdui-parser port (#17645) #19398 / card Bump .objectui-sha — the console pin is 584 commits behind objectui main #17429 ✓; pair PR chore(objectui): bump the console pin to 87af769e9a3e with 27 re-measured citations, and fold in the two-code sdui-parser port (#17645) #19398 / card [finding] the sdui-parser lockstep has drifted again since the record was taken: objectui carries 26 diagnostic codes, this copy 24 — inert-quick-add and member-type-mismatch exist only on the renderer side #17645 UNJUDGED — the card's newest claim (5753263180) puts Branch: inside the Claim: sentence, which parses to zero branches, so the gate fell back to the older os-try-charles claim naming claude/issue-17645-sdui-lockstep-port. CI's "The card this PR closes must claim this branch" is success and the PR touches no governed surface, so this does not bind the review; the remedy the gate prints is one comment on [finding] the sdui-parser lockstep has drifted again since the record was taken: objectui carries 26 diagnostic codes, this copy 24 — inert-quick-add and member-type-mismatch exist only on the renderer side #17645 with Branch: \claude/issue-17429-objectui-pin-bump`on a line of its own. The same gate reads the PR body'sClause-②: yesline as NEAR-MISS/describing (prose after the value); a bareClause-②: yes` line in the body would make it a clean carrier.

Implemented-by: claude/issue-17429-objectui-pin-bump
Reviewed-by: session_012GcsUbuqFGBibkEDMRC1eE

VERDICT: PASS

---

Generated by Claude Code

[os-project-manager]

Provenance — director seat, summon #25 (session_012GcsUbuqFGBibkEDMRC1eE), 2026-09-21T00:17Z

• Contract review of record: 5753767727, Served-tier: CONTRACT_REVIEW_TIER, judged head fc25a5925d7d6b55ed25b81d5f1fcef4be836109 (the current head), VERDICT: PASS — seven ① judgments RIGHT, none WRONG, none NOT MEASURED; the reviewer's transcript re-read by this seat: 328 of 328 requests stamped at the tier constant, zero fallbacks. Independence pair Implemented-by: claude/issue-17429-objectui-pin-bump / Reviewed-by: session_012GcsUbuqFGBibkEDMRC1eE (the dev rounds were mode:subagent under this seat; the reviewer was an isolated subagent whose verdict this seat adopts verbatim).
• What the reviewer re-derived first-hand: the pin and its ancestry; the manifest hash; all eight fold-in blobs SAME as the port branch; check:sdui-lockstep 0 with 26 = 26 codes from git archive extracts of both trees; the three ablations reproduced to the PR's exact counts; 21 of the 27 citation rows and all 7 content assertions re-read at both objectui shas, both SUBSTANCE-REVERSED describes checked against the code at the new pin; check-issue-citations 0 and #10274 probed DELETED first-hand; diff file set = PR file set; check:generated 15/15 current after a real build; console built through the lock (VERDICT 0, dist stamp 87af769e9a3e, check:console-sha 0); all three changeset levels judged right.
• needs:contract-review stripped in this stroke from this PR, Bump .objectui-sha — the console pin is 584 commits behind objectui main #17429 and [finding] the sdui-parser lockstep has drifted again since the record was taken: objectui carries 26 diagnostic codes, this copy 24 — inert-quick-add and member-type-mismatch exist only on the renderer side #17645 (label-write, read back). check-clause2-carriers --pair 19398: the PR / Bump .objectui-sha — the console pin is 584 commits behind objectui main #17429 pair reads clean; the PR / [finding] the sdui-parser lockstep has drifted again since the record was taken: objectui carries 26 diagnostic codes, this copy 24 — inert-quick-add and member-type-mismatch exist only on the renderer side #17645 pair could not be judged mechanically — row C9 wants a Release: from the offline domain:devx holder itself, and this seat's reclaim Release: (5753791571's predecessor on that thread) is by a different author — so the body now carries [finding] the sdui-parser lockstep has drifted again since the record was taken: objectui carries 26 diagnostic codes, this copy 24 — inert-quick-add and member-type-mismatch exist only on the renderer side #17645 as Part of, and the seat closes [finding] the sdui-parser lockstep has drifted again since the record was taken: objectui carries 26 diagnostic codes, this copy 24 — inert-quick-add and member-type-mismatch exist only on the renderer side #17645 with the landing record after the merge. The take-over is the maintainer's instruction (「18723 os-try-charles 已下班，你也接手」), recorded on the card.
• CI on the head at the reviewer's read and at this seat's own read 2026-09-21T00:10Z: all 42 check runs green, the seven required contexts success, Check Changeset success.
• Landing: non-governed diff (no governed path, 1551 lines changed, under the 5000 line) ⇒ the seat lands it: ready for review and auto-merge (squash) through the queue in this stroke. PR sdui-parser: port inert-quick-add and member-type-mismatch from objectui, re-record the lockstep and bump the pin #18723 was closed superseded earlier (5753281831).

---

Generated by Claude Code