[finding] the ViewNavigationConfig parity pin authors navigation.view in four places — the key objectstack#18619 retires as an ADR-0049 tombstone, so the twin needs the tombstone pin #9667

Activity

1. os-bill

   added

   finding

   on Sep 17, 2026
2. os-tesla commented on Sep 18, 2026

   

   os-tesla

   on Sep 18, 2026 – with Claude

   Collaborator

   More actions

   Population note for whoever takes this card — ⛔ not a re-grade, ⛔ not a claim. Seat domain:ui#2, session_018HrVaotisyhgmot9o2MLRq.

   This card names the four declaration sites in packages/types/src/__tests__/view-navigation-config-spec-parity.test.ts. A census run by the objectui#9874 dev, re-measured at source by this seat on origin/main 00c4df509 at 2026-09-18T18:45Z, finds the retired key authored at five more sites in three other files — each of which types never on the bump to @objectstack/spec ≥ 17.5.0. Printed, ⛔ not counted:

   packages/types/src/__tests__/gantt-declared-keys.test.ts:117   navigation: { mode: 'page' as const, view: 'task_detail', openNewTab: false },
   packages/types/src/__tests__/gantt-declared-keys.test.ts:191   navigation: { mode: 'page', view: 'task_detail', openNewTab: false },
   packages/types/src/__tests__/object-view-unmirrored-keys-7779.test.ts:386   navigation: { mode: 'drawer', view: 'summary_view' },
   packages/plugin-list/src/__tests__/ListView.ganttNavigationForward-7334.test.tsx:183   navigation: { mode: 'split', view: 'leave_summary', size: 'lg' },
   packages/plugin-list/src/__tests__/ListView.ganttNavigationForward-7334.test.tsx:185   expect(node.navigation).toEqual({ mode: 'split', view: 'leave_summary', size: 'lg' });
   

   ⚠️ :185 is an assertion over the same shape rather than a fresh authoring site; it breaks on the bump for the same reason and is listed so the next reader does not have to rediscover it.

   ⭐ In the same command, this seat's probe shows three authoring lines in this card's own file (:133 :134 :135) — the card's 「four places」 is its own count and is ⛔ left as filed; the discrepancy is recorded rather than resolved, because a recount is the claiming seat's to do with its own instrument.

   ⚠️ Two more facts the taker should have, both measured on the objectui#9874 branch and now on main:

   • PR objectui#9936 (card objectui#9874) inverted two existing pins that asserted the retired substitution as intended behaviour — useNavigationOverlay.modeDefault and gridNavigationMembers-8071 — keeping their authored view fixtures, because those fixtures are the one input the old and new implementations disagree about. ⇒ those two fixtures are in the same population as the five above.
   • The published doc comment half is filed separately as objectui#9938 (packages/types/src/objectql.ts:2546 teaches the retired key as legal authored metadata). ⛔ Not folded in here: a doc comment an author reads and a test fixture fail differently.

   ⛔ This seat changed no label, no grade and no assignee on this card; it remains pm:queue, unassigned.

   _{domain:ui#2 execution seat · session_018HrVaotisyhgmot9o2MLRq · readings above taken in this act on origin/main 00c4df5}

   ---

   Generated by Claude Code
3. os-tesla

   added

   pm:blocking

   pm:dispatched

   and removed

   pm:queue

   on Sep 19, 2026
4. os-tesla

   self-assigned this

   on Sep 19, 2026
5. os-tesla commented on Sep 19, 2026

   

   os-tesla

   on Sep 19, 2026 – with Claude

   Last edited by os-tesla

   Collaborator

   More actions

   Claim: objectui#9667 — domain:ui#2 execution seat

   Seat: domain:ui#2
   Session: session_018HrVaotisyhgmot9o2MLRq
   Branch: claude/issue-9667-navigation-view-tombstone
   Thread-read: 5734641478

   Claimed at 2026-09-19T01:52Z. Labels first, this comment second, the thread re-read third — two comments (triage's grading, and this seat's own population note), read to the last page.

   ⭐ 取卡前置 — North Star clause 3, re-taken for THIS claim

   product repo (open-issue enumeration, PRs excluded · 取数时刻 2026-09-19T01:51:48Z)	open P0	open P1
   objectstack-ai/objectstack	4	47
   objectstack-ai/objectui	0	6

   ⇒ gate CLOSED. This card is priority:p2 on its own label, and it passes by the clause's second half — 「解锁产品 P0/P1 的仪器卡沿链继承其优先级」. It is now the sole Blocked-by: target of objectui#9860 (priority:p1), whose delivery PR objectui#9877 is HELD on Spec Main Shape Gate being red. ⇒ it inherits P1 along the chain. ⛔ The priority:p2 label is left as triage set it; the inheritance is a dispatch-order fact, not a re-grade, and ⛔ this seat does not re-grade.

   1. ⭐ The gate now names this card's exact lines — the trigger triage asked for has fired

   Triage's own warning was 「⚠️⚠️ 次序是本卡最要紧的一格 … ⛔ 本卡不得早于该 PR 落地」. Both halves of that ordering are now satisfied, measured in this act:

   • objectstack PR objectstack#18619 is MERGED — read on the API: merged: true, 2026-09-17T10:06:57Z, merge commit 3d8779d6c5. ⇒ the contract has moved; a tombstone pin written now is ⛔ not ahead of its subject.
   • The retirement is live and measured from the outside. Spec Main Shape Gate check run 105816840308 on PR objectui#9877's synced head 88c0b1d56, 2026-09-19T01:37:53Z → 2026-09-19T01:42:30Z, compiled this repo against @objectstack/spec built from objectstack 8b8258d910db and reported ⛔ 6 diagnostics in 2 files, every one of them TS2322: Type 'string' is not assignable to type 'undefined' — which is what an authored view: '…' becomes once the key leaves the union. Five of the six are this card's own :123 :133 :134 :135 :146. The sixth is gantt-declared-keys.test.ts:191, already recorded on this card in this seat's population note of 2026-09-18T18:45Z.

   ⇒ this is no longer a card filed ahead of a break. The break is here, it is the only thing keeping a P1 card's PR red, and this card is its whole subject.

   2. Population re-measured in THIS act, and ⭐ why the gate's six is not the population's eleven

   Read on objectui origin/main 1db82373f767c5331d7ce8d1bfb9beb4420f7cf5 (freshly fetched at claim time). ⚠️ The card body's line numbers ⛔ have not moved — :123 :133 :134 :135 :146 are still exact — but the card's 「four places」 counts three declarations plus the annotated one; the reading below is by line.

   site	in the gate's six?	why
   types/…/view-navigation-config-spec-parity.test.ts :123 :133 :134 :135 :146	yes	typed declarations — excess/incompatible property checking fires
   types/…/gantt-declared-keys.test.ts :191	yes	annotated ObjectGanttSchemaTS
   types/…/gantt-declared-keys.test.ts :117	no	const good = { … } — an un-annotated literal, so no check fires. ⭐ A reading, not a mystery
   types/…/object-view-unmirrored-keys-7779.test.ts :386	no	same shape suspected, ⛔ not confirmed by this seat
   plugin-list/…/ListView.ganttNavigationForward-7334.test.tsx :183 :185	no	@object-ui/plugin-list is one of the 3 tasks that never ran — turbo printed Tasks: 23 successful, 26 total, Failed: @object-ui/types#type-check
   types/src/objectql.ts :2546	n/a	a published docblock, filed separately as objectui#9938

   ⇒ ⭐ fixing only the six the gate printed would move the fail-fast one package down and cost another whole cycle. The dispatch carries the packages/types + packages/plugin-list population instead. That the gate's count is a lower bound is objectui#9944's property, ⛔ not a surprise.

   3. ⛔ One population is deliberately EXCLUDED, and the reason is a state, not a judgement

   PR objectui#9936 (card objectui#9874) inverted two existing pins — useNavigationOverlay.modeDefault and gridNavigationMembers-8071 — and kept their authored view fixtures on purpose, because those fixtures are the one input the old and new implementations disagree about. objectui#9874 is in the decision box (needs-user-decision).

   ⇒ ⛔ a dev does not edit the fixtures an open decision card is about. Those two files are out of this card's face and the dispatch says so by name. ⚠️ Consequence stated in advance: the gate may report them once @object-ui/react actually runs, and that will be objectui#9874's to answer, ⛔ not a failure of this dispatch.

   4. ⚠️ The hazard this card walks into — named now so the dev does not discover it as a surprise

   The card's ask is 「assert that navigation.view is refused」. This repo's convention for that is @ts-expect-error — widely used across packages/types/src/__tests__/. ⭐ But this package resolves @objectstack/spec@17.4.0, where the key still exists, so an @ts-expect-error written today would be unused and fail the local type-check. ⇒ the tombstone half may not be writable until the dependency moves, while dropping the authorings is valid on both the installed 17.4.0 and on spec@main and is what clears the gate.

   ⇒ the dispatch splits the card's ask accordingly and instructs a measured stop on the half that cannot be written yet, rather than a guess. ⛔ This seat does ⛔ not assert that the tombstone pin is impossible — it asserts that nobody has measured it, and names who must.

   5. In-flight and file-face

   This seat's other in-flight cards and faces: objectui#9955 → packages/core/src/utils/sort-query.ts, its test, and packages/plugin-grid/src/__tests__/gridArrayArmOrderby-8973.test.tsx; objectui#9952 → packages/app-shell/src/views/metadata-admin/inspectors/ConditionBuilder.tsx. ⛔ Both disjoint from this card's face — no serialization owed.

   packages/types/src/__tests__/view-navigation-config-spec-parity.test.ts
   packages/types/src/__tests__/gantt-declared-keys.test.ts
   packages/types/src/__tests__/object-view-unmirrored-keys-7779.test.ts
   packages/plugin-list/src/__tests__/ListView.ganttNavigationForward-7334.test.tsx
   .changeset/*
   

   6. Prior rulings

   The governing decision is named on the card and re-read here: director batch #126 item 4, executed by objectstack PR objectstack#18619 (card objectstack#16885), whose execution item 3 reads verbatim 「the contract twins gain the tombstone pin」. ⇒ this card is execution of a standing ruling, ⛔ not a question — which is why it sits in the queue and ⛔ not in the decision box.

   Next: dispatch to one os-dev subagent, one worktree.

   ---

   Generated by Claude Code
6. os-tesla commented on Sep 19, 2026

   

   os-tesla

   on Sep 19, 2026 – with Claude

   Collaborator

   More actions

   Dispatch — one os-dev, one worktree

   Issued at 2026-09-19T01:53Z by domain:ui#2 (session_018HrVaotisyhgmot9o2MLRq). Base at issue time: objectui origin/main 1db82373f767c5331d7ce8d1bfb9beb4420f7cf5. ⛔ Branch from a freshly fetched origin/main and re-take every reading below against YOUR base.

   ⭐ This card is the sole Blocked-by: target of objectui#9860 (priority:p1), whose delivery PR objectui#9877 is HELD on a red gate. Everything below is shaped by that: the job is to get Spec Main Shape Gate further, not merely to close the six lines it printed.

   ⭐ Provenance — two labels, never mixed

   • [SEAT-MEASURED] — read at source by this seat in the claim act, on 1db82373f.
   • [CARD-CLAIM] — the filing seat's or triage's reading, relayed. ⛔ Re-measure before building on it.

   1. What is wrong

   [SEAT-MEASURED] objectstack PR objectstack#18619 (card objectstack#16885, director batch #126 item 4) retired ListViewSchema.navigation.view as an ADR-0049 tombstone and merged 2026-09-17T10:06:57Z (merged: true read on the API in the claim act). This repo's twin still authors that key. Once the key leaves the union its type is undefined, so every authored view: '…' is TS2322: Type 'string' is not assignable to type 'undefined'.

   [SEAT-MEASURED] Spec Main Shape Gate check run 105816840308, compiled against @objectstack/spec built from objectstack 8b8258d910db, reported 6 diagnostics in 2 files — all of them this key.

   2. ⭐ The population — ⛔ fix all of it, not the six the gate printed

   [SEAT-MEASURED] on 1db82373f, with the reason each site is or is not in the gate's six:

   packages/types/src/__tests__/view-navigation-config-spec-parity.test.ts
     :123  const navigation: ViewNavigationConfig = { view: 'summary_view' };          ← in the six
     :133  const grid:  Pick<ObjectGridSchema,  'navigation'> = { navigation: { view: 'summary_view' } };   ← in the six
     :134  const view:  Pick<ObjectViewSchema,  'navigation'> = { navigation: { view: 'summary_view' } };   ← in the six
     :135  const named: Pick<NamedListView,     'navigation'> = { navigation: { view: 'summary_view' } };   ← in the six
     :146  view: 'summary_view',            (inside the "every spec key" literal)      ← in the six
     :15   docblock sentence teaching that `{ view: 'summary_view' }` is legal authored metadata
   
   packages/types/src/__tests__/gantt-declared-keys.test.ts
     :191  navigation: { mode: 'page', view: 'task_detail', openNewTab: false },        ← in the six (annotated `ObjectGanttSchemaTS`)
     :117  navigation: { mode: 'page' as const, view: 'task_detail', openNewTab: false },  ← NOT in the six
   
   packages/types/src/__tests__/object-view-unmirrored-keys-7779.test.ts
     :386  navigation: { mode: 'drawer', view: 'summary_view' },                        ← NOT in the six
   
   packages/plugin-list/src/__tests__/ListView.ganttNavigationForward-7334.test.tsx
     :183  navigation: { mode: 'split', view: 'leave_summary', size: 'lg' },            ← NOT in the six
     :185  expect(node.navigation).toEqual({ mode: 'split', view: 'leave_summary', size: 'lg' });   ← NOT in the six, and an ASSERTION not an authoring
   

   ⭐ [SEAT-MEASURED] :117 is absent from the gate's six because it sits in const good = { … } — an un-annotated literal, so no excess/incompatible-property check fires on it. That is a reading with a cause, ⛔ not a site the gate cleared. object-view-unmirrored-keys-7779.test.ts:386 is suspected to be the same shape — ⛔ this seat did not confirm that; confirm it yourself and say which it is.

   [SEAT-MEASURED] @object-ui/plugin-list is absent because it never ran: turbo printed Tasks: 23 successful, 26 total with Failed: @object-ui/types#type-check. ⇒ fixing only packages/types moves the fail-fast one package down and costs another cycle. ⛔ Do not do that.

   3. ⭐ The two halves of the ask, and why they may not land together

   The card asks for two things: drop the authorings, and pin the tombstone (assert the key is refused).

   [SEAT-MEASURED] The hazard, named so you do not discover it as a surprise: this package resolves @objectstack/spec@17.4.0, where navigation.view still exists. This repo's convention for a tombstone pin is @ts-expect-error — used widely across packages/types/src/__tests__/. An @ts-expect-error written today would therefore be unused, and an unused one is itself an error. ⇒ the tombstone pin may not be writable until the dependency moves; dropping the authorings is valid on both 17.4.0 and spec@main.

   So:

   • Half A — drop/repair the authorings. Required, and it is what clears the gate. Each site keeps its own purpose. :123's case is 「accepts the spec-valid config that omits mode」, so it needs a still-live non-mode key — preventNavigation, openNewTab, size and width are all still on the type; pick one and keep the test's point intact. ⛔ Do not weaken a case into vacuity to make it compile; if a case's whole subject was view, say so and propose what it becomes.
   • Half B — the tombstone pin. MEASURE whether it can be written against the installed 17.4.0. If it can, write it the way this repo's sibling retirements do (read two or three of them first — action-callback-retired-7068, ai-insights-retired-8800, bare-kanban-node-key-retired-8802 are in the same directory) and say which convention you followed. If it cannot, ⛔ do not fake it and ⛔ do not leave it silent: say so in the PR with the exact reading that shows it, and note in the changeset that the pin is owed when the dependency moves. A measured 「half B is not writable yet」 is a good outcome.

   4. ⛔ Excluded by name — a state, not a judgement

   [CARD-CLAIM, this seat's own earlier note] PR objectui#9936 (card objectui#9874) inverted two pins and kept their authored view fixtures deliberately, because those fixtures are the one input the old and new implementations disagree about:

   packages/react/src/hooks/__tests__/useNavigationOverlay.modeDefault.test.tsx
   packages/plugin-grid/src/__tests__/gridNavigationMembers-8071.test.tsx
   

   objectui#9874 is needs-user-decision — in the decision box. ⛔ A dev does not edit the fixtures an open decision card is about. Both files are outside this face. ⚠️ The gate may report them once @object-ui/react runs; that is objectui#9874's to answer and ⛔ not yours.

   ⛔ Also excluded: packages/types/src/objectql.ts:2546, the published docblock teaching the retired key — filed separately as objectui#9938. A doc comment an author reads and a test fixture fail differently. ⛔ Do not fold it in. ⚠️ But view-navigation-config-spec-parity.test.ts:15 is yours: it is a docblock inside a file you are already rewriting, and leaving it teaching the retired key while the code below stops using it is the contradiction this whole family is about.

   5. Tests

   ⭐ The parity file's premise is Equal<ViewNavigationConfig, NavigationConfig> — the objectui type IS the spec type, re-exported by reference since objectui#4588. ⛔ Do not weaken that assertion to make anything compile; it is the file's reason to exist.

   [CARD-CLAIM] The card reports a lit control in that file: ViewNavigationConfig occurs on 20 lines, against a fabricated token at 0. Re-take it; it is your cheap check that the file is the one you think it is.

   Every change needs a control that proves the file still tests something: a key that is still live must still be accepted at each site you touch.

   6. Declared file face

   packages/types/src/__tests__/view-navigation-config-spec-parity.test.ts
   packages/types/src/__tests__/gantt-declared-keys.test.ts
   packages/types/src/__tests__/object-view-unmirrored-keys-7779.test.ts
   packages/plugin-list/src/__tests__/ListView.ganttNavigationForward-7334.test.tsx
   .changeset/*
   

   ⛔ Anything outside it is a stop-and-report, ⛔ never a rider. [SEAT-MEASURED] No open PR touches these four files.

   7. Stop conditions — a stop with evidence is a good outcome

   • A site's whole subject is view, so repairing it would leave a vacuous case ⇒ stop and report with a proposal.
   • Half B is not writable against 17.4.0 ⇒ ⛔ not a stop for the whole card: land half A and report half B with the reading.
   • object-view-unmirrored-keys-7779.test.ts:386 turns out to need something structurally different ⇒ report it rather than improvising.
   • You find the retirement is not live on your base (an authored view still type-checks against spec@main) ⇒ stop immediately; the premise has moved and this card would be ahead of its subject again.

   8. Deliverable

   Branch, code, tests, changeset, push, draft PR with Fixes #9667. ⛔ Do not flip out of draft, ⛔ do not enqueue, ⛔ do not merge. Run the repo's fast checks before pushing. Report back as structured JSON.

   ⚠️ Your local type-check runs against @objectstack/spec@17.4.0, where the key still exists — so a green local run does not prove the gate goes green. Say so in your report rather than implying coverage you do not have; the gate re-runs on objectui#9877 after this lands and that is the real reading.

   In the PR body: the per-site before/after, which sites were and were not in the gate's six and why, your finding on :386, and half B's outcome with its evidence.

   ---

   Generated by Claude Code
7. os-tesla commented on Sep 19, 2026

   

   os-tesla

   on Sep 19, 2026 – with Claude

   Collaborator

   More actions

   os-dev-report

   {
     "issue": 9667,
     "status": "done",
     "branch": "claude/issue-9667-navigation-view-tombstone",
     "pr": "https://github.com/objectstack-ai/objectui/pull/9986",
     "session": "session_018HrVaotisyhgmot9o2MLRq",
     "premise_still_valid": true,
     "summary": "Half A landed: every authoring of the retired `navigation.view` key is gone from the four declared-face files, each site keeping its own subject on a still-live key (`preventNavigation` for the mode-less cases, `mode`+`openNewTab` for the gantt legs, `mode`+`size` for the object-view fixture and the ListView forwarding case). No case was weakened into vacuity and the parity file's `Equal` premise, its four per-interface pins and `_KeysAreExactlyTheSpecSix` are untouched. `SpecDeclaredKeys` deliberately still lists `view`: a `retiredKey()` tombstone is `z.never().optional()`, so the key stays DECLARED while only its authoring type collapses to `undefined` — a comment now says so, because dropping it would redden that assertion rather than repair it. Half B is measured NOT writable and is reported, not faked. Assignee was already set by the PM (os-tesla); this seat wrote none.",
     "tests": "All heavy runs serialized behind scripts/pm/os-verify-lock.sh, slot objectui-9667; exit codes captured by redirecting to a file and reading $? before any pipe. GREEN: `pnpm --filter @object-ui/types run type-check` exit 0; `pnpm --filter '@object-ui/plugin-list^...' run build` (dependency closure, required because plugin-list's type-check reads sibling dist typings) exit 0; `pnpm --filter @object-ui/plugin-list run type-check` exit 0; `pnpm exec vitest run` over the four touched files = Test Files 4 passed (4), Tests 239 passed (239); `pnpm exec vitest run packages/types/ packages/plugin-list/` = Test Files 288 passed, Tests 5813 passed. GATES exit 0: check:control-bytes (8053 files), check:new-line-citations (0 new citations — and two pre-existing cross-file path:line citations were removed from the docblock this PR rewrites), check:test-path-roots, check:changeset-claims, check:pending-changeset-literals, check-changeset-presence.mjs (4 source files of 2 released packages, 1 changeset, EMPTY frontmatter = declared as releasing nothing), check-changeset-no-major.mjs. ESLINT (declared narrowing, three pieces of evidence): --no-inline-config --format json over the 4 files = 4 files linted, 0 errors, 5 warnings, all no-explicit-any on pre-existing harness lines; universe read from eslint.config.js's own test-file blocks; type-aware linting is NOT enabled (no project/projectService parser option), so this diff cannot move any untouched file's verdict. Repo-wide `pnpm lint` is CI's run. HALF-B MEASUREMENT (the discriminating one): `@ts-expect-error` written over an authored `view: 'summary_view'`, then `pnpm --filter @object-ui/types type-check` printed exactly one diagnostic — `view-navigation-config-spec-parity.test.ts(200,7): error TS2578: Unused '@ts-expect-error' directive.`, Exit status 2 — against the installed @objectstack/spec@17.4.0 whose NavigationConfigSchema still declares view as an optional string. Probe removed and its absence verified by grep. PLUGIN-LIST ABLATION (subject + lit control, direction predicted before running, both legs restored from HEAD and the restore proven by blob-hash comparison against `git rev-parse HEAD:PATH`, not by a trap's exit code): SUBJECT — `mode: 'split'` mutated to `mode: 99999` at the repaired site, type-check exit 0 with ZERO diagnostics; LIT CONTROL — a plain `const x: number = 'nope'` added to the same file, type-check exit 2 with TS2322. On-disk mutation proven by anchor-match-count-equals-1 plus grep counts of injected and removed text, never by an editor's exit code. NO ABLATION for half A, and the reason is stated rather than skipped: the repair is a deletion whose failure mode only exists on a spec this checkout cannot resolve, so a mutation leg would be green on both sides and would prove nothing. NOT MEASURED: Spec Main Shape Gate. A green local run does NOT prove the gate goes green — this package resolves @objectstack/spec@17.4.0 where navigation.view still exists and an authored name still compiles. The gate's re-run on objectui#9877 is the only reading that answers it. NOT MEASURED: CI on PR objectui#9986, in_progress at report time and not waited on.",
     "mcp_calls": "0 — no MCP GitHub tool was called, read or write.",
     "api_writes": "2 REST proxy writes: POST /repos/objectstack-ai/objectui/pulls (draft PR 9986); POST /repos/objectstack-ai/objectui/issues/9667/comments (this report). NO label write: in this repo labels are applied by the `Auto Label PRs` workflow (.github/workflows/labeler.yml, actions/labeler, config .github/labeler.yml), and it already put ['package: types', 'plugin', 'tests'] on PR 9986 — read back and compared: that set is exactly what the config's rules yield for these paths, since plugin-* packages are covered by the `plugin` family label by design rather than a per-package one. Nothing was stripped and nothing is owed. The `skip-changeset` label was deliberately NOT applied: in this repo it is inert and the empty-frontmatter changeset is the authoritative declaration. Also: 2 `git push` (the empty-branch routing probe, then the commit). One POST was rejected by the agent proxy for a missing Content-Type header before reaching GitHub — no write occurred — and was re-sent with the header.",
     "open_questions": [],
     "out_of_scope_findings": [
       "to file (3 classes, dedupe words): class (c) — `packages/plugin-gantt/README.md` is PUBLISHED (that package's files[] lists README.md) and positively instructs authors: \"To choose which detail view opens, use the declared `view` member (a form-view name, e.g. summary_view)\". That is the retired key, so the README now teaches metadata the runtime REFUSES at parse with the tombstone's prescription. Distinct from objectui#9938, which owns the packages/types/src/objectql.ts docblock: different package, and this one is an INSTRUCTION rather than an explanation of a past collapse. Outside this card's declared file face, so not ridden in. Handoff: the same family as objectui#9938; no open PR touches packages/plugin-gantt/README.md. Dedupe words: plugin-gantt README navigation.view | summary_view | navigation.view tombstone | ADR-0049 sibling half | objectui#9938",
       "noted, not filed: the dispatch's [SEAT-MEASURED] reason for plugin-list's two sites being absent from the gate's six — \"it never ran\" — is incomplete. Measured with a subject/control pair: the harness types its schema parameter as a Record keyed by string with `any` values, so no property check fires and the sites would NOT have appeared even had the task run. Recorded in the PR body rather than filed, because it is a correction to a reading, not a defect. Handoff: the next PR to touch ListView.ganttNavigationForward-7334.test.tsx; the file now carries the reading in a comment.",
       "noted, not filed: three `safeParse` legs in object-view-unmirrored-keys-7779.test.ts authored the retired key and asserted it is ACCEPTED. They were NOT in the seat's enumeration and are invisible to tsc, but a tombstoned key refuses at the PARSE, so they would have gone red at RUNTIME on the dependency bump. Same defect, same declared-face file, different instrument — repaired in this PR rather than filed. Handoff: none needed; it is in the diff.",
       "noted, not filed: this PR's first docblock draft cited a sibling retirement pin by name, and that pin asserts its own retired spelling appears nowhere in the tracked tree by `git grep` — so the citation reddened it. Caught by the full-package run, citation changed, re-run green. A test file's prose is inside other tests' denominators. Handoff: none; no defect remains, recorded so the next author of a retirement docblock does not rediscover it."
     ]
   }

   ---

   Generated by Claude Code
8. os-tesla commented on Sep 19, 2026

   

   os-tesla

   on Sep 19, 2026 – with Claude

   Collaborator

   More actions

   Seat review — ACCEPT half A, ACCEPT the measured refusal of half B. PR objectui#9986 lands when CI converges.

   Seat domain:ui#2, session_018HrVaotisyhgmot9o2MLRq, 2026-09-19T02:18Z. Dev report status: "done", open_questions: [].

   1. ⭐ Half B — the measurement this dispatch asked for instead of a guess, and it came back NO

   The dispatch said the tombstone pin 「may not be writable until the dependency moves」 and ordered it measured, ⛔ not assumed. The dev measured it and reported the discriminating line:

   @ts-expect-error written over an authored view: 'summary_view', then pnpm --filter @object-ui/types type-check printed exactly one diagnostic — view-navigation-config-spec-parity.test.ts(200,7): error TS2578: Unused '@ts-expect-error' directive., exit 2

   ⇒ correct, and the reason is worth stating in one line because it is the whole shape of the problem: an @ts-expect-error is an assertion that something FAILS, so it is red in exactly the world where the tombstone has not arrived. The pin is owed the day this package's @objectstack/spec moves, and ⛔ it is not written now in a disabled or conditional form — 「a pin that passes on both sides of the contract it guards is the phantom check this file exists to prevent」. The file records the one command that re-derives whether it is writable, and ⛔ records no version number to go stale against. That is the right shape.

   2. ⭐ The catch that would have been the easy mistake

   SpecDeclaredKeys still lists view, deliberately, and the diff says why:

   A retiredKey() tombstone is z.never().optional(): the key is still IN the object shape, so keyof still yields it — only its input type collapses from string | undefined to undefined. DECLARED and WRITABLE are two different facts, and this alias measures the first one.

   ⇒ dropping view from that list would redden _KeysAreExactlyTheSpecSix against both the installed spec and spec@main. ⭐ Getting that backwards is the obvious "tidy" a later reader will attempt; the file now argues against it in place. The Equal< ViewNavigationConfig, NavigationConfig > premise and the four per-interface pins are untouched — I checked, that was the file's reason to exist and the dispatch fenced it.

   3. ⛔⛔ A CORRECTION THIS SEAT OWES — my enumeration was a lower bound produced by a filter I never declared

   The dev repaired three safeParse legs in object-view-unmirrored-keys-7779.test.ts (:820, :823, :826–:827) that authored the retired key and asserted it is ACCEPTED. They were not in my population table. They are invisible to tsc — but a tombstoned key refuses at the parse, so they would have gone red at RUNTIME on the dependency bump.

   Re-derived in this act, so the cause is measured and not guessed: my enumeration grepped view: ' and then filtered to lines that also contain navigation. Those safeParse lines read slot.safeParse({ view: 'summary_view' }) — the container is not named on the line, so my own filter excluded them silently. ⭐ A co-occurrence filter on one line is not a population query, and I published its output as one.

   A value-keyed enumeration (summary_view · task_detail · leave_summary) run in this act returns the true set. Two carriers it finds that are ⛔ outside this card's face:

   • packages/plugin-gantt/README.md:54–:62 — published (files: ["dist","README.md",…]) and it instructs 「To choose which detail view opens, use the declared view member」, then explains in the next sentence how an undeclared key rejects the whole config. ⇒ filed as objectui#9987.
   • packages/react/src/hooks/useNavigationOverlay.ts:31 — ⛔ CLEAN, checked in this act: it is the tombstone's own account of past behaviour, and :37–:40 state the replacement route by name. ⛔ Nobody should re-open it.

   4. ⛔⛔ A SECOND CORRECTION — my 「plugin-list never ran」 reason was incomplete, and it changes what a reader should expect

   I wrote, marked [SEAT-MEASURED], that ListView.ganttNavigationForward-7334.test.tsx's two sites were absent from the gate's six because @object-ui/plugin-list never ran. The dev measured that this is not the reason, and I re-derived it at source:

   packages/plugin-list/src/__tests__/ListView.ganttNavigationForward-7334.test.tsx
    :127  async function viewProps(schema: Record<string, any>) {
    :139  const viewSchema = async (schema: Record<string, any>) => (await viewProps(schema)).schema;
    :183      navigation: { mode: 'split', view: 'leave_summary', size: 'lg' },   ← passes through `Record<string, any>`
   

   ⇒ no property check ever fires on that literal. Those two sites would not have appeared in the gate's six even if the task had run. The task not running is true; it is not the cause. Full correction and its consequence for the gate are on objectui#9860, where a reader would act on it.

   5. Verification and landing

   ⛔ NOT governed — re-run on the FINAL file list in this act. Declared face honoured: 4 files + changeset, ⛔ nothing outside it. The dev's own readings, relayed and attributed: @object-ui/types and @object-ui/plugin-list type-check exit 0; 288 test files / 5813 tests green across both packages; the four touched files 239 tests green; gates all exit 0.

   ⭐ And the ablation is declared honestly in the one place it could have been faked: no ablation for half A, with the reason stated rather than skipped — 「the repair is a deletion whose failure mode only exists on a spec this checkout cannot resolve, so a mutation leg would be green on both sides and would prove nothing」. A refused measurement with its reason beats a decorative one. The plugin-list leg DID get a subject-plus-lit-control ablation, restore proven by blob hash against git rev-parse HEAD:PATH and ⛔ not by a trap's exit code.

   ⚠️ Spec Main Shape Gate is NOT MEASURED by this PR and the dev says so plainly: this package resolves @objectstack/spec@17.4.0, where an authored name still compiles. The gate's re-run on PR objectui#9877 after this lands is the only reading that answers it — and that is this card's whole purpose.

   Implemented-by: os-dev subagent (branch claude/issue-9667-navigation-view-tombstone)
   Reviewed-by: domain:ui#2 — the diff review above plus two re-derivations that corrected this seat's own published readings

   _{domain:ui#2 execution seat · session_018HrVaotisyhgmot9o2MLRq · the PR file list, the parity and unmirrored-keys patches, the Record<string, any> harness lines, the plugin-gantt README quotation with its files[] entry, the useNavigationOverlay.ts exclusion and the value-keyed re-enumeration were all read by this seat in this act on objectui origin/main 1db82373f767c5331d7ce8d1bfb9beb4420f7cf5; the type-check, test and TS2578 readings are the dev's, relayed and attributed.}

   ---

   Generated by Claude Code
9. os-tesla

   closed this as completedin #9986on Sep 19, 2026
10. os-tesla commented on Sep 19, 2026

    

    os-tesla

    on Sep 19, 2026 – with Claude

    Collaborator

    More actions

    Landed — verified by content on origin/main

    PR objectui#9986 merged 2026-09-19T02:58:39Z; squash commit c255b38384384175249f2811c41155d3d03577ca. Reading taken at 2026-09-19T03:08Z against a freshly fetched origin/main at that tip. BEFORE = f156329bb5474c2f847210aa3066dde8de5e5c9c.

    ⛔ Verified by CONTENT on both refs, never by sha and never by ancestry. Subject rows are one git grep -c pair summed over the four declared-face files.

    reading	BEFORE	AFTER
    authored retired view names (summary_view · task_detail · leave_summary) across the four files	16	2
    preventNavigation — the replacement carrier	3	15
    CONTROL — SpecDeclaredKeys still lists 'view'	1	1 — unmoved
    CONTROL — the Equal< ViewNavigationConfig, NavigationConfig > premise	2	2 — unmoved

    ⚠️ ⭐ The 16 → 2 row needs its remainder read, ⛔ not assumed to be two leftover sites. Both survivors are prose inside comments that record the retirement:

    view-navigation-config-spec-parity.test.ts:21   * mode-less config as `{ view: 'summary_view' }`, and `view` has since been
    view-navigation-config-spec-parity.test.ts:168  // was `{ view: 'summary_view' }`; `view` is a tombstone since
    

    ⇒ zero authorings remain; two historical mentions stay, which is the right kind of record. A bare count of 2 would have read as unfinished work — the same case-artefact shape this seat has been bitten by twice.

    ⭐ The two controls are the ones that could have gone wrong, and both held

    SpecDeclaredKeys still lists 'view' on both refs, deliberately. A retiredKey() tombstone is z.never().optional(): the key stays in the object shape, so keyof still yields it — only its input type collapses from string | undefined to undefined. DECLARED and WRITABLE are two different facts. Dropping view from that list would have reddened _KeysAreExactlyTheSpecSix against both the installed spec and spec@main, ⛔ not repaired it. That is the "tidy" a later reader will attempt, and the file now argues against it in place.

    And the Equal< ViewNavigationConfig, NavigationConfig > premise is unmoved at 2 — the file's whole reason to exist, fenced by the dispatch and honoured.

    Half B — refused with a measurement, ⛔ not faked

    The tombstone pin objectstack#18619's execution item 3 asks for is not writable from here yet, and the dev measured it rather than guessing:

    @ts-expect-error over an authored view: 'summary_view' ⇒ view-navigation-config-spec-parity.test.ts(200,7): error TS2578: Unused '@ts-expect-error' directive., exit 2

    ⇒ an @ts-expect-error is an assertion that something FAILS, so it is red in exactly the world where the tombstone has not arrived. The pin is owed the day this package's @objectstack/spec moves; it is ⛔ not written now in a disabled or conditional form, because 「a pin that passes on both sides of the contract it guards is the phantom check this file exists to prevent」. The file records the one command that re-derives whether it is writable, and ⛔ records no version number to go stale against.

    What this unblocks

    This card was the sole Blocked-by: of objectui#9860 (priority:p1), whose delivery PR objectui#9877 is HELD on Spec Main Shape Gate. That PR's branch has been synced onto this tip (88c0b1d56 → 27ade033b, behind_by 4 → 0, zero-file-write) and the gate is re-running as check run 105829004143. ⚠️ Its result is the only reading that answers whether these six diagnostics are gone — a green local run here proves nothing, because this package resolves @objectstack/spec@17.4.0 where an authored name still compiles. The dev stated that limitation itself rather than implying coverage it did not have.

    ⚠️ Two corrections this seat published alongside this card, both to its own readings: the population table was a lower bound produced by an undeclared single-line co-occurrence filter (three safeParse legs escaped it, repaired in this PR), and the 「plugin-list never ran」 reason was incomplete — that harness types its schema parameter Record<string, any>, so those sites would not have appeared even had the task run. Full text on objectui#9860 at 5738559450.

    Closing state: auto-closed completed by the PR's Fixes line. pm:dispatched and the assignee cleared in the same label write; pm:blocking comes off with them — objectui#9860's edge is discharged.

    Implemented-by: os-dev subagent (branch claude/issue-9667-navigation-view-tombstone)
    Reviewed-by: domain:ui#2 — the review at 5738556418 and the content verification above

    ---

    Generated by Claude Code
11. os-tesla

    removed

    pm:dispatched

    pm:blocking

    on Sep 19, 2026
12. os-tesla

    removed their assignment

    on Sep 19, 2026
13. akarma-synetal

    added a commit that references this issue on Sep 28, 2026

    test(types,plugin-list): drop the retired `navigation.view` authoring…

    ...

    This issue will close once commit c255b38 is merged into the 'main' branch.

    c255b38