finding(types/plugin-kanban): an authored kanban document can PASS safeValidateSchema and still render empty — the validator and the renderer honour different faces

Filed by the domain:ui PM seat from the objectui#7645 flight. The dev seat surfaced it and deliberately did not file it, because it sits against objectui#7651's open maintainer decision and objectui#6172's settled ruling — filing across a live maintainer decision is the PM's call. Recorded by the repository itself rather than inferred.

The hazard

For an authored node of type 'kanban', two different types are authoritative depending on who is asking:

asker	honours	where
the validator — `safeValidateSchema`, i.e. what the CLI's `validate` / `check` commands accept	`DeclarativeKanbanSchema`	`packages/types/src/complex.ts`'s own docblock states it is the `'kanban'` arm of `ComplexSchema` → `AnyComponentSchema`
the renderer — `ObjectKanbanRenderer`, registered for the `'kanban'` key	`@object-ui/plugin-kanban`'s own `KanbanSchema`	`packages/plugin-kanban/src/index.tsx`

The two are structurally unrelated dialects: the declarative face carries draggable, column color, card labels / assignees / priority; the plugin face carries objectName, groupBy, cardTitle, cardFields, card badges and thirteen more.

⇒ A board can pass validation and render empty. An author runs objectui validate, gets a green result, ships it, and the renderer reads none of the keys the validator just blessed.

⚠️ This is NOT the latent class

objectui#7645 was a false claim in a type map with zero indexed accesses — nothing consumed it, so nothing broke. This is different: both consumers are live, both run on real authored documents, and they disagree. A green validate is exactly the signal an author would trust.

Not a duplicate of the neighbours

objectui#6172 ruled that both faces are kept (option A, 2026-08-31), and PR refactor(types)!: rename the declarative Kanban trio, leaving the bare names to plugin-kanban #7643 landed the rename that makes them distinguishable. It did not rule which one an authored document answers to — that question was not before it.
objectui#7645 / PR fix(types): SchemaRegistry's kanban entry stops describing a component it cannot name #7662 made SchemaRegistry['kanban'] stop naming either face. That removed a false type claim; it did not reconcile the two consumers.
objectui#7651 asks whether ObjectKanban should gain a real record-source ladder. Adjacent, different axis.

The decision needed

Which face is authoritative for an authored kanban document? Everything else follows from it:

if the plugin face is authoritative, the validator arm is validating the wrong shape and should be reconciled to it;
if the declarative face is authoritative, the renderer registered for that key does not honour the published contract;
if both are legitimately authorable, the validator needs to accept either and something has to tell an author which one they wrote.

⛔ Not dispatchable until that is answered — every remedy is a different piece of work, and two of the three move a published surface.

Refs: objectui#7645 (origin) · objectui#6172 (both faces kept) · objectui#7651 (open decision) · PR #7643 · PR #7662.

Generated by Claude Code

Maintainer ruling recorded — (a): for an authored `type: 'kanban'` document the plugin dialect is authoritative. `@object-ui/types`' `'kanban'` arm (type + Zod mirror) is rewritten to declare exactly the shape `ObjectKanbanRenderer` reads (`objectName` / `groupBy` / `cardTitle` / `cardFields` / card `badges` … — the 18-member `KanbanSchema` `plugin-kanban` consumes), and `plugin-kanban` conforms to that declaration; the `DeclarativeKanban*` trio loses its only live consumer and retires under ADR-0049

Director seat, summon #14, session session_01LsEjuNMPitCHwEfYftZ1um (GitHub os-warren), 2026-09-05. Provenance: maintainer, live PM chat, decision batch #41 (item 4, presented with the recommendation (a) and the explicit note that it reverses the "keep both faces" half of the objectui#6172 ruling), verbatim reply 「同意」. Premise: the card's table — the validator (safeValidateSchema, CLI validate / check) honours DeclarativeKanbanSchema, the renderer registered for 'kanban' honours the plugin's KanbanSchema; PR #7643's census: all four registered kanban renderers consume the plugin dialect and none consumes the declarative one; objectui#6086 measured the consequence (a validated board renders empty).

Ruled: (a). One contract for one type key: the 'kanban' arm of ComplexSchema → AnyComponentSchema → safeValidateSchema accepts what the registered renderer reads and refuses what it does not. Dependency direction stays correct — @object-ui/types declares, @object-ui/plugin-kanban imports and conforms (it already depends on types). The DeclarativeKanbanSchema / DeclarativeKanbanColumn / DeclarativeKanbanCard types and their three Zod mirrors, renamed by PR #7643, are retired in the same change: objectui#6172's own stop condition was "if the renamed copy has no retained value, escalate" and the retained value it cited was precisely the validator arm — this ruling removes that value, so the retirement is the ruled outcome, not a reversal by a seat. Not taken: (b) (renderer rewritten to the static columns/cards dialect — four renderers rewritten and the data-bound board lost), (c) (both authorable, a new static renderer under a new key — a renderer nobody asked for; ④).

Why (① ≥50%): a validator and a renderer bound to the same type key must read the same contract or a green validate means nothing; (a) is the only shape where declared = rendered. ② every existing consumer is the data-bound board; the static board has zero. ③ "passes validation, renders empty" is the worst feedback an AI author can get; (a) turns it into a named refusal. ④ one dialect fewer.

Execution: domain:ui seat, M, packages/types (complex.ts, zod/complex.zod.ts, index.ts, zod/index.zod.ts, registry.ts) + packages/plugin-kanban (import the declared type instead of its local one; behaviour unchanged) + the zod-mirror-parity, handler-key and one-authority ratchets re-keyed; the docs page api/schema-reference.md's kanban example moves to the plugin dialect (#6973 is folded or closed by this — the seat decides and links it). Pins: an objectName/groupBy board passes safeValidateSchema; a columns/cards board is refused with a message naming the retired shape; the four renderers' props still type-check against the declared schema. Clause-②: yes ⇒ needs:contract-review on the PR, parked for the director. Changeset: @object-ui/types minor with a BREAKING banner — the 'kanban' validator arm's accept set is replaced and the six DeclarativeKanban* exports (renamed by #7643 in the same release line) are retired, stated as the second step of that rename; @object-ui/plugin-kanban patch. Sequencing: after #7651 (ruled B in the same batch — no data ladder, so data?: any[] stays a raw-row input in the declared shape) and clear of PR #7662's SchemaRegistry change (already merged per the card).

State transition, same stroke: needs-user-decision → pm:queue. finding · domain:ui unchanged. Ledger: objectstack #12708, batch #41. Related: objectui#6172 · #7645 · #7651 · #6086 · #6973 · PR #7643 · PR #7662.

Generated by Claude Code

Sequencing correction before this card is dispatched — PR #7662 has not merged

The ruling's Execution paragraph reads:

Sequencing: after #7651 (ruled B in the same batch …) and clear of PR #7662's SchemaRegistry change (already merged per the card).

The first half holds — #7651 is closed not_planned (2026-09-05T02:09:54Z). The second half does not. Read from the API just now:

	reading
PR #7662 state	open, draft, `merged: false`
mergeable_state	`behind`
gate	carrying a second contract-review REFUSE (comment 5548793740)

So registry.ts does not yet carry BaseSchema & { type: 'kanban' } on main; it still carries DeclarativeKanbanSchema. An executor starting from the ruling's premise would edit a line that is not there yet, and would contend with #7662 on the same file.

Surfaced by the independent contract review of #7662 (flag F5), which had to read both states to grade that PR — not a defect in the ruling's substance, only in one parenthetical about a sibling PR's status. The ruling's disposition is unaffected: either way, (a) re-points this entry.

Disposition from the domain:ui seat: this card is NOT dispatched yet. It is held until PR #7662 lands, for two reasons that are the ruling's own:

Both changes edit packages/types/src/registry.ts, and fix(types): SchemaRegistry's kanban entry stops describing a component it cannot name #7662's outstanding remediation touches the comment at :201 that (a) will rewrite.
fix(types): SchemaRegistry's kanban entry stops describing a component it cannot name #7662's entry is the transitional state under (a). Landing them out of order means the executor writes the final state onto a file whose intermediate state is still in review, and the review record for fix(types): SchemaRegistry's kanban entry stops describing a component it cannot name #7662 stops describing any tree that ever existed.

⛔ Not an indefinite park. #7662's refusal is wording-only and its remediation is dispatched; this card moves the moment that PR is merged. It stays pm:queue so it is visible in the pool meanwhile.

One more note for whoever executes it: the ruling folds or closes #6973 at the seat's discretion. #6973 was flagged earlier in this seat's log as possibly resting on a stale premise — re-measure it before folding, do not inherit the assumption.

Generated by Claude Code

✅ Hold lifted — PR #7662 has merged. This card is dispatchable now.

I held this card on 2026-09-05T02:39Z (comment 5548799174) because the ruling's sequencing note said PR #7662 was "already merged per the card" and it was not. That is now resolved by the fact, not by a re-reading:

PR #7662 merged 2026-09-05T07:14:56Z as bc640ec56 (merged: true read from the API, Fixes #7645, card closed). It landed after three contract-review rounds; the third returned PASS.

The exact state ruling (a)'s executor inherits — read from `main`, not from a report

origin/main bc640ec56, packages/types/src/registry.ts:

'kanban': BaseSchema & { type: 'kanban' };

⛔ So do not start from DeclarativeKanbanSchema. The entry is the transitional form, and (a) re-points this one line at the declared type.

The #7662 contract review enumerated the rest of the inheritance, and judged it does not pre-empt any part of (a):

the one type line above;
the BaseSchema import at registry.ts:9, which becomes unused after the re-point — a tidy, not a gate (noUnusedLocals: false, @typescript-eslint/no-unused-vars is warn);
the comment block at registry.ts:187-217, whose "⛔ Do not restore … without moving the renderer's dialect into a layer this package may depend on" names (a)'s route rather than forbidding it;
both pin test files, which import DeclarativeKanbanSchema and therefore retire with the trio — disclosed in their own docblocks. The reviewer flagged _KeyKept as worth carrying forward.

Two numbers in the ruling that the executor should NOT inherit

The ruling says "the 18-member KanbanSchema". The measured body count is 19 (AST census, controls lit). Re-derive rather than trusting either number.
The ruling folds or closes objectui#6973 at the seat's discretion. finding(docs): schema-reference.md's KanbanSchema example mixes both dialects — cards from the plugin copy with color / draggable from the types copy, so it matches no declaration #6973 was flagged in this seat's log as possibly resting on a stale premise — re-measure it before folding, do not inherit the assumption.

Also cleared since the ruling was written

[Decision] Should ObjectKanban get a real record-source ladder? — KanbanSchema.data is any[], so it has no data binding to resolve #7651 — closed not_planned (ruled B in the same batch). The fence it named is gone.
finding(types): three more type names are declared twice with disagreeing shapes — FormField, MarkdownSchema, KanbanSchema — and one doc asserts FormField is declared once #6172 — closed completed by PR refactor(types)!: rename the declarative Kanban trio, leaving the bare names to plugin-kanban #7643. Its "keep both faces" half is what (a) reverses; the ruling says so explicitly.

Card stays pm:queue, unassigned. Clause-②: yes per the ruling, so whoever takes it owes a draft PR with needs:contract-review on both carriers, and a minor changeset with a BREAKING banner.

Generated by Claude Code

Claim: PM loop round 25 — domain:ui execution seat
Session: session_01KbJQ1y1J12nZxYzFWhP8Q3
Branch: claude/issue-7664-kanban-plugin-dialect-authoritative
Worktree: objectui-issue-7664
Domain: domain:ui
File surface: packages/types/src/complex.ts, packages/types/src/zod/complex.zod.ts, packages/types/src/index.ts, packages/types/src/zod/index.zod.ts, packages/types/src/registry.ts, packages/plugin-kanban/**, the zod-mirror-parity / handler-key / one-authority ratchets, content/docs/api/schema-reference.md (stop on breach; explain in the report)
Container & model: M, mode:subagent, model: fable — node scripts/pm/dispatch-gates.mjs --tier run this round returns no path-derived mandate; the fable tier is the clause-② CONTENT mandate, and the gate's own line calls itself a floor, never a clearance.
Clause-②: yes
Serial constraints cleared: Two live contentions, both declared rather than waited on. #7731 (claude/issue-7433-parity-population-derived) edits only packages/types/src/__tests__/zod-mirror-parity.test.ts + a changeset — the exact ratchet this card must re-key, and it is rewriting how that population is DERIVED. #7726 (claude/issue-7530-predicate-envelope-declared) edits packages/types/src/index.ts, zod/index.zod.ts, zod/base.zod.ts and the same parity test. Neither is a blocker: merge the base branch, ⛔ never rebase. ⚠️ If #7731 lands first the parity population becomes derived rather than listed — re-read that file before writing any ledger count, do not port a hand-count. Also in flight on packages/types: #7732 (complex.ts ChatMessage region — a different interface from this card's kanban arm), #7730 (zod/base.zod.ts only). Cleared by measurement, not memory: #7651 closed not_planned, #7662 merged bc640ec56 — the ruling's own sequencing note said #7662 was already merged when it was not, corrected on this card at comment 5548799174; it has since become true. registry.ts on main now reads 'kanban': BaseSchema & { type: 'kanban' }.

Generated by Claude Code

os-dev-report

{
  "issue": 7664,
  "status": "done",
  "branch": "claude/issue-7664-kanban-plugin-dialect-authoritative",
  "pr": "https://github.com/objectstack-ai/objectui/pull/7743",
  "premise_still_valid": true,
  "summary": "Ruling (a) executed: @object-ui/types' 'kanban' arm now declares the plugin dialect (KanbanSchema / KanbanColumn / KanbanCard / CardTemplate / ColumnWidthConfig, member for member the plugin's shape) with five Zod mirrors applied by safeValidateSchema, SchemaRegistry['kanban'] names it, plugin-kanban/src/types.ts re-exports it (behaviour unchanged), and the six DeclarativeKanban* exports plus both 7645 pins retire under ADR-0049 with a 7664 pin on each side. The retired dialect's inert keys (board draggable, column color — zero read sites) are ?: never tombstones refused by name naming DeclarativeKanbanSchema; onColumnAdd / onCardAdd carry over; the parity, handler-key and docs-retired ratchets are re-keyed (parity on top of PR #7731's derived population, merged in: EXPECTED_MIRROR_PAIRS = 159); schema-reference.md moves to the plugin dialect; changesets types minor (BREAKING banner) + plugin-kanban patch. Re-derived numbers: 19 live members + 3 tombstones (ruling said 18; `type` is the difference); objectui#6973 was already closed completed at 2026-09-05T00:43Z, so nothing to fold — referenced, not closed. Deviations to review: the pin 'a columns/cards board is refused' was read as 'a board in the retired dialect is refused' because the ruled shape itself declares columns[].cards[] (the catalog's static boards render); the two 'could not run' gates (readme-exports, doc-snippets) need a whole-workspace build and are left to CI. Dispatch inheritance held: the card carried assignee os-sam and pm:dispatched at dispatch; the label set on the card was never written by this seat. Landed since BASE and merged (no rebase): #7726, #7730, #7731, #7732, #7686.",
  "tests": "All at bd1fc7111 (head; union re-run after the merge commit). Lock VERDICT command-exit 0 for: pnpm --filter @object-ui/types build; tsc -p packages/types/tsconfig.test.json --noEmit; tsc -p packages/plugin-kanban/tsconfig.json --noEmit; tsc -p packages/plugin-kanban/tsconfig.test.json --noEmit => ALL_TSC_GREEN (--listFiles: both test projects include the new pin files, the three re-keyed ratchets, complex.ts, complex.zod.ts; plugin project reads types/dist/complex.d.ts + registry.d.ts). pnpm exec vitest run --maxWorkers=2 over 10 files (types 7664 pin, zod-mirror-parity, handler-keys-6124, docs-retired-7340, phase2-schemas; plugin 7664 pin, registration; catalog kanban-column-cards-6939, safe-validate-corpus-6318; scripts one-authority-6273) => 'Test Files 10 passed (10)', 'Tests 389 passed (389)'. Ablation A: registry.ts 'kanban': KanbanSchema -> CalendarViewSchema, mutation proved on disk (anchored count 1->0 / 0->1), tsc types test project exit 2 with exactly 1 error in kanban-plugin-dialect-authoritative-7664.test.ts; restore via git checkout HEAD -- (absolute path, trap), proved by git hash-object == HEAD blob 3b5ce94b and empty git diff HEAD --stat; no rebuild needed, the project reads src. Ablation B: complex.zod.ts draggable refusal -> z.boolean().optional(), mutation proved (1->0 / 0->1), vitest on the pin exit 1 '2 failed | 9 passed' (the two refusal tests); restore proved by hash == HEAD blob b5a341dd and empty diff; no rebuild in this leg, vitest resolves @object-ui/types through the source alias so dist was not on the path. Gates exit 0 with verdict lines: check:phantom-deps, check:doc-types, check:doc-fences, check:control-bytes (6309 files), check:unreferenced-sources, check:spec-symbols, check:published-dist, check:element-data-source-declaration, check:sdui-registration-pins, check:self-import, check:published-tsconfig-exclude, scripts/check-doc-links.mjs ('Links are valid across 17 scan roots'), check-changeset-presence (2 changesets for 14 source files of 2 packages), check-changeset-no-major. ESLint narrowed: eslint --format json over the 16 changed lintable files => 0 errors, 48 warnings (no-explicit-any on members carried verbatim); population evidence: repo lint scripts are 'eslint .' per package (no --no-inline-config — that spelling reddens 4 untouched files on main), JSON-counted run over the touched directories 356 files 0 errors, eslint.config.js has no parserOptions.project / type-checked config so untouched verdicts cannot move. NOT MEASURED locally: check:readme-exports ('312 self-import(s) could not be judged … run pnpm build first') and check:doc-snippets (exit 2 'I could not run') — both need every package's dist; left to CI. CI status at report time: in_progress (draft PR just opened).",
  "mcp_calls": "2 — two directed search_issues dedupe queries (both non-empty, self-validating); every other read/write went through git, the public web payload, or repo-scoped REST (issue create, PR create, label add, comment)",
  "open_questions": [
    {
      "question": "The ruling's refusal pin says 'a columns/cards board is refused', but the ruled shape (the plugin's KanbanSchema) itself declares columns[].cards[] and the catalog's static boards render every card. Which reading should the contract review hold?",
      "options": [
        "A — as implemented: refuse the retired dialect at its distinguishing inert keys (board draggable, column color) by name; retired card keys (labels/priority/dueDate/…) stay accepted because a card is an open record and those are legitimate field names",
        "B — additionally refuse the retired card keys by name (would refuse real records whose fields are named priority/dueDate/labels; bucketCardsIntoColumns pushes raw records into lanes)",
        "C — refuse every static columns/cards board (contradicts the declared type, breaks the two catalog entries and kanban-column-cards-6939)"
      ],
      "recommendation": "A, because it is the only reading where the mirror refuses what the renderer does not read and accepts what it does; B refuses data, C refuses the declaration"
    }
  ],
  "out_of_scope_findings": [
    "filed as #7742: the ruled KanbanSchema carries three zero-read members (allowCollapse, cardTemplates, columnWidths) and ObjectKanban reads an undeclared titleField fallback for cardTitle; root README's kanban example authors columns[].value/label/color, a column shape no renderer reads — each an ADR-0049 enforce-or-remove decision on the shape this PR declared (no assignee, no label)",
    "already open, not re-filed: #7322 (the sibling object-kanban arm requires groupField with zero read sites while the same renderer reads groupBy) and #7712 (plugin-kanban reads schema.filter undeclared) — both surfaced by the same read census; referenced from #7742 and the PR body"
  ]
}

Generated by Claude Code

Contract review (clause ②), fresh and full after the R1 REFUSE — PASS — PR #7743 at head `1a5915f2`

Director seat, summon #15, session_01TezFG8ZMrNH6n5VTNpPpdH (os-zhuang), 2026-09-05T16:2xZ, on the maintainer's instruction 「然后执行契约复审」. Tier fuse read by get_session this session: session_context.model = last_served_model = CONTRACT_REVIEW_TIER. This is a new review of the whole 25-file remediated diff at 1a5915f2, not a re-read of the REFUSE (5552425067) — that verdict's findings were used only as an attack list. Readings are this seat's own, taken from the PR head fetched as refs/pull/7743/head and from objectui origin/main 737037f2.

Implemented-by: session_01KbJQ1y1J12nZxYzFWhP8Q3's os-dev round (dev reports 5552339519, 5552854017 on this card; branch claude/issue-7664-kanban-plugin-dialect-authoritative)
Reviewed-by: session_01TezFG8ZMrNH6n5VTNpPpdH

① Derived judgments — accept set and public face

#	claim	this seat's reading	verdict
R1	`onCardClick` is a declared runtime slot on the `'kanban'` arm: callable on the TS face, refused by name on the mirror	PR head `packages/types/src/zod/complex.zod.ts:152`: `onCardClick: handlerKeyRefusal('onCardClick', 'runtime-slot', 'Card click handler')` beside `onCardMove` (:151) and `onQuickAdd` (:155) — a declared refusing arm, so `BaseSchema`'s `.passthrough()` no longer keeps the value. TS face `complex.ts`: `onCardClick?: (card: KanbanCard, event?: unknown) => void`. The widening the REFUSE measured is closed at the source, not only at the probe.	correct
R1-disp	runtime-slot rather than retired	objectui `origin/main` `plugin-kanban/src/index.tsx:194` forwards `onCardClick={schema.onCardClick}` in the same block as the two kept slots; `ObjectKanban.tsx:151` declares an `onCardClick` prop and `:938-940` its substitute calls the authored one. An authored `onCardClick` on a `type: 'kanban'` document runs ⇒ live on every channel measured; `retired` would have declared a live key dead.	correct
2	Ledger re-keyed: `RUNTIME_SLOT` 44 → 45, `ALL_SITES` 66 → 67, `RETIRED` 22 unchanged; `KeepsFunction<KanbanSchema['onCardClick']>` type assertion	Diff of `handler-keys-json-refusal-6124.test.ts`: `toHaveLength(45)` / `toHaveLength(22)` / `new Set(ids).size).toBe(67)`, the `onCardClick` row re-keyed to `KanbanSchema`, and the `Expect<KeepsFunction<…>>` line present. The dev's ablation (`?: never` back in ⇒ TS2344 at that line + TS2322 in `zod-mirror-parity`) shows the type instrument is lit.	correct
3	The `'kanban'` arm declares the plugin dialect (19 live members + `onCardClick` + 3 tombstones `draggable` / `onColumnAdd` / `onCardAdd`; column `color?: never`); `registry.ts` `'kanban': KanbanSchema`; six `DeclarativeKanban*` exports removed	Mirror hunk read: `KanbanSchema = BaseSchema.extend({...})` with `draggable: retiredDeclarativeKanbanKey(...)`, `onColumnAdd` / `onCardAdd` as `'retired'` arms; `index.ts` exports `KanbanCard`, `KanbanColumn`, `KanbanSchema`. Whole-name removal for the six exports is right — a whole exported name has no carrier to tombstone, and the retirement is the ruling's own outcome (objectui#6172's stop condition).	correct
4	No remaining consumer of the retired names	`git grep DeclarativeKanban` on objectui / cloud / hotcrm `origin/main` outside the PR's own files: two historical `.changeset` entries and one comment line in `scripts/check-doc-links.mjs:508` (no code reference). Zero live importers.	correct
5	`plugin-kanban/src/types.ts` becomes a re-export of the five declared types; behaviour unchanged	Files list −218/+31 on that file, `ObjectKanban.tsx` import becomes `import type`; the plugin's public type names are preserved through the re-export.	correct
6	Refusal pin re-read as A (refuse `draggable` and column `color`, not `columns[].cards[]`)	The plugin dialect declares `columns[].cards[]` and `bucketCardsIntoColumns` pushes raw records into `cards`, so refusing card field names would refuse the declaration itself; A executes the ruling's principle (accept what the registered renderer reads). The REFUSE reviewer reached the same answer independently.	correct

② semver

@object-ui/types minor with a BREAKING banner (validator arm's accept set replaced; six exports retired as the second step of the #7643 rename) and the accept-set bullet for all five handler arms added in remediation; @object-ui/plugin-kanban patch (re-exports; the "member for member" claim corrected to "plus four"). check-changeset-no-major green. Matches the ruling (5548643216).

③ Boundary flags

REFUSE's non-blocking (a)(b)(c): all three addressed in the remediation — plugin changeset wording, tombstone JSDoc citing prong 2 of the discriminator, schema-reference.md documents onCardClick again and the 7340 control grows to 7.
bind correction (a live BaseSchema member, not an undeclared read) carried; finding(types,plugin-kanban): the ruled KanbanSchema carries three zero-read members (allowCollapse, cardTemplates, columnWidths) and the board reads an undeclared titleField — enforce-or-remove on the shape objectui#7664 declared #7742 extended with five re-measured rows by comment, not a new card.
Systemic finding filed as objectui#7753 (a hand-written handler-key ledger cannot see a deletion; the new derived guard covers 3 of 67 sites). Correctly its own card.
Dev question on member count (18 vs 19): re-derived by AST, pinned; the ruling's number was short by type. Confirmed.

Evidence and landing checks

CI on 1a5915f2: 29 success / 3 skipped / 0 red (individually enumerated by the seat at 15:43Z; re-read here). mergeable_state: clean against base 83c77dc3.
⚠️ Read before the queue runs: origin/main has moved past the merge-base with three packages/types/src commits (feat(types): refuse chartType on a chart series by name, pointing at type #7737 chartType refusal, docs(types): align two layout @default docblocks with the renderer fallbacks #7736 docblocks, feat(types)!: retire ChatbotSchema.displayMode as an ADR-0049 tombstone; remove the Display Mode control and seed (objectui#7654) #7739 ChatbotSchema.displayMode retirement). No textual conflict, but the count ratchets (EXPECTED_MIRROR_PAIRS = 159, the 7340 docs census 26, RETIRED 22) may need re-deriving in the merge-queue build if any of those three moved a count. That is a CI/ratchet matter for the dispatching seat, not a contract finding; it does not hold this PASS.
Governed: check-governed-merges --test on the 25 paths ⇒ 0 hits, ordinary queue landing. content/docs/releases/: 0. Fixes #7664 is correct — this card's whole scope lands with the PR.

Clearing, same stroke: needs:contract-review comes off this card and PR #7743 with a provenance comment citing this PASS; then ready + auto-merge (SQUASH).

Generated by Claude Code

Landing blocked at the queue — merge lap owed on PR #7743 (conflict with #7758), `domain:ui` seat

Director seat, session_01TezFG8ZMrNH6n5VTNpPpdH, 2026-09-05T16:4xZ. Measured on the timeline and the tree, not inferred:

PR feat(types): the kanban arm declares the plugin dialect; the DeclarativeKanban* trio retires (objectui#7664) #7743 added_to_merge_queue 16:14:25Z → removed_from_merge_queue 16:17:28Z; mergeable_state now dirty.
git merge-tree --write-tree --name-only origin/main origin/pr-7743 (objectui origin/main 4dfdcc3c): one conflict, packages/types/src/__tests__/zod-mirror-parity.test.ts. Cause: test(types): watch the mirror-wider-than-declared direction with a fifth parity ledger #7758 (test(types): watch the mirror-wider-than-declared direction with a fifth parity ledger) landed after this PR's merge-base 83c77dc3 and edits the same ledger this PR re-keys (EXPECTED_MIRROR_PAIRS = 159, KnownDrift, SPEC_DERIVED_PAIRS). Also on main since the merge-base: fix(types): bind the ListView exportOptions mirror to the spec field; drop the ListView casts (#6956) #7763 (ListView exportOptions mirror), feat(types)!: retire ChatbotSchema.displayMode as an ADR-0049 tombstone; remove the Display Mode control and seed (objectui#7654) #7739 (ChatbotSchema.displayMode tombstone), feat(types): refuse chartType on a chart series by name, pointing at type #7737 (chartType refusal), docs(types): align two layout @default docblocks with the renderer fallbacks #7736 — any of which can move a count this PR pins. This is the ratchet risk the PASS (5553091448) named.

Handover to the domain:ui seat (session_01KbJQ1y1J12nZxYzFWhP8Q3) — a patch round on the original dev, not REWORK: merge origin/main into claude/issue-7664-kanban-plugin-dialect-authoritative with a merge commit (⛔ no rebase / amend / force-push), resolve the parity-ledger conflict by composing both edits (#7758's fifth ledger + this PR's kanban re-key; re-derive EXPECTED_MIRROR_PAIRS from the population, do not pick a side), re-run the ratchet instrument (tsc -p packages/types/tsconfig.test.json --noEmit) and the types/plugin-kanban vitest, and push. The card stays pm:dispatched; nothing else on this card changes.

Contract review after the lap: the head will move by a merge commit plus a conflict resolution in a test ledger. Post the new head here and this seat re-reads the resolution and extends the PASS to it (a merge-only move with a ledger re-derivation is an extension, not a new review); only then does auto-merge go back on. ⛔ Do not re-arm before that comment.

Generated by Claude Code

Nudge — the merge lap on PR #7743 is still owed (domain:ui seat, session_01KbJQ1y1J12nZxYzFWhP8Q3 / assignee os-sam). At 17:32Z the branch tip is still 1a5915f2 (15:18Z), mergeable_state unresolved, no comment since the handover 5553249289 (16:4xZ). The PR sits out of the merge queue until the conflict in packages/types/src/__tests__/zod-mirror-parity.test.ts (vs #7758) is composed on a merge commit and pushed. Nothing else blocks landing: contract PASS 5553091448 on record, CI green on the current head, governed test 0 of 25. When the new head is pushed, post it here; this seat extends the PASS and re-arms auto-merge. Director seat, session_01TezFG8ZMrNH6n5VTNpPpdH; reported to the maintainer.

Generated by Claude Code

Contract review (clause ②) — PASS 5553091448 extended to head `b115f08d`; gate cleared on both carriers

Director seat, summon #15, session_01TezFG8ZMrNH6n5VTNpPpdH (os-zhuang), 2026-09-05T19:2xZ. The re-hang by the domain:ui seat (5554184339) was the right act under 「head 后移或无结论才重挂」; this is the extension it was waiting for, read from the tree, not from the reports.

Delta 1a5915f2 → b115f08d, read in full:

commit	content	reading
`2702b741`, `724a7a5e`, `b5de2f20`	three merges of `origin/main` (up to `2fcefb95`)	`git diff origin/main origin/pr-7743` shows the PR's own delta unchanged: the parity file's net change vs main is exactly the reviewed re-key — `DeclarativeKanban*` → `KanbanCardSchema` / `KanbanColumnSchema` / `KanbanSchema` (+ `CardTemplateSchema`, `ColumnWidthConfigSchema`) in imports, `MIRRORS`, `SPEC_DERIVED_PAIRS`; `KnownDrift` `KanbanSchema: onCardMove
`b115f08d`	`zod-mirror-parity.test.ts` −4 / +0: the two `WiderThanDeclared` rows `'complex.zod.ts#DeclarativeKanbanColumnSchema': 'cards'` and `'complex.zod.ts#DeclarativeKanbanSchema': 'columns'` with their `SCHEMA-NODE` doc lines	Correct and the only correct remedy: both rows key on pairs this PR retires, so on the merged tree they fail `assertionWiderLedgerKeysAreRegistered` (TS2344) — a sequencing consequence of #7758 and this PR landing in order, not a contract change. Re-keying them to the new pairs was measured `WiderOf = never` (the drift they recorded does not exist on the plugin-dialect arm), so there is nothing to carry. Coverage lost: none. The ledger's own remedy 4 ("a key DISAPPEARED — correct or delete the entry").

Ratchet evidence on the new head: tsc -p packages/types/tsconfig.test.json --noEmit exit 0 (was exit 2 with exactly the one TS2344 before the deletion); lit control — mutating CarouselSchema: 'items' turns two assertions red; RUNTIME_SLOT 45 / RETIRED 22 / ids 67 / EXPECTED_MIRROR_PAIRS 159 re-measured on b115f08d, all equal to the figures the PASS was taken on. CI on b115f08d: 29 success / 3 skipped / 0 red, individually enumerated by the seat and re-read here. mergeable_state: clean.

Two readings worth carrying past this PR (the seat's, confirmed here): (1) the merge lap's first stop-and-report was correct — a row in another PR's ledger is a fence, and the ruling to delete came from the dispatching seat with the measurement, not from the dev's preference; (2) vitest green while tsc red — the parity ledger is compile-time; tsc -p packages/types/tsconfig.test.json is the instrument, never the vitest run.

Verdict: PASS extended to b115f08d. Every derived judgment in 5553091448 holds on this head; the accept set, public face and semver are unchanged. Clearing needs:contract-review from this card and PR #7743 in the same stroke; then auto-merge (SQUASH). If the head moves again before the queue takes it, this extension does not carry — re-hang per the rule.

Implemented-by: session_01KbJQ1y1J12nZxYzFWhP8Q3 (dev reports 5553898255, 5554084984)
Reviewed-by: session_01TezFG8ZMrNH6n5VTNpPpdH

Generated by Claude Code

落地确认：PR #7743 已合并。本卡由 Fixes 自动关闭。三条判据全过，读数如下。

domain:ui PM 席（session_01KbJQ1y1J12nZxYzFWhP8Q3，round 6）。

① 合并读数 —— 读 `merged` 字段本身

项	读数
`merged`	`true`
`merged_at`	`2026-09-05T19:39:34Z`
`merge_commit_sha`	`2af1fa7c03d5edf311872870e896435436ccc5f2`
`merged_by`	`os-zhuang`（经合并队列）
diff	25 files, +1470 / −601
本卡	`closed_at 2026-09-05T19:39:36Z`

⛔ 没有把 mergeable_state 当合并证据（合并后它读 unknown）。

② 队列轨迹 —— 读 timeline 事件，⛔ 不读 `auto_merge`

added_to_merge_queue     16:14:25Z  os-zhuang
removed_from_merge_queue 16:17:28Z  github-merge-queue[bot]   ← MERGE_CONFLICT 弹出
added_to_merge_queue     19:24:57Z  os-zhuang                  ← PASS 延展后重新武装
merged                   19:39:34Z

中间那次弹出正是三轮 merge lap 的起点：main 在此期间移动了九个提交，最后一次是 PR #7782（同样改 zod-mirror-parity.test.ts），逼出第三次 merge b5de2f206，之后才是 PM 裁定 A 的 4 行删除 b115f08d0。

③ 关闭列表 —— 只关了本卡

PR 正文里的闭合关键词只有一个：7664。正文另提到 17 个编号（#6124 #6172 #7712 #7753 等）全部走 Refs / 散文，没有一个带闭合关键词。⇒ 无误伤。

⚠️ 后续载体存活核对：10 个全部 open

卡	状态
#7744 · #7745	open，`pm:queue`
#7747	open，`pm:dispatched`（本轮已派）
#7750 · #7751	open，`pm:queue`（`security`，p2）
#7753 · #7756 · #7765	open —— ⚠️ *三张此前一个 `pm:` 标都没有**，见下
#7786	open，`pm:dispatched`（本轮已派）
#7787	open，`pm:blocked`（`Blocked-by: #7652`）

⚠️ 本席自己漏的三张，已补

#7753 / #7756 / #7765 此前是半状态 —— 卡活着，但对任何按 pm:* 扫描的派发视图不可见（#7753、#7756 一个标都没有；#7765 只有 finding）。三张都是本会话自己立的，漏标是本席的账。已补：

finding(types): the #6124 handler-key ledger is hand-written by name, so it cannot see a DELETED key — 64 of its 67 sites have no read-site derivation #7753 finding tests package: types domain:ui p2 pm:queue —— finding(types): 28 zod-mirror keys are declared z.function(), which NO JSON document can satisfy — 14 in-repo corpus files author them and are silently invalid #6124 的 handler-key 台账按名字手写，因此看不见一个被删掉的键。BaseSchema 是 .passthrough()，键离开 arm 不是被拒绝，而是不再被判断、值被保留；67 个站点里 64 个没有读点派生。⇒ 与本卡同一个家族的下一次，且每一道门都会是绿的。
finding(plugin-detail): RelatedList.lookupLabelResolution renders inside a waitFor predicate — a 50ms lookup-batch delay exhausts the V8 heap and kills the whole vitest worker #7756 finding tests plugin domain:ui p2 —— RelatedList.lookupLabelResolution 在 waitFor 谓词里渲染；lookup 批延迟多 50ms 就耗尽 V8 堆并杀死整个 vitest worker（tests 0ms，什么都没报出来）。
The network-escape guard prescribes a stubGlobal + unstubAllGlobals pair that races a component re-read started after the test body returns #7765 finding tests domain:ui p2 —— 网络逃逸守卫自己打印的 Fix: 建议（stubGlobal + unstubAllGlobals）正是造成 flaky(app-shell): RecordDetailView.approvalDeclaredActions trips the network-escape guard intermittently — the afterEach stub teardown races a still-pending effect #7439 那次 flake 的配对。⇒ 一道门在教人写出它自己会抓的东西。

⇒ 三条判据全过，后续载体零丢失，三张半状态已修复。

Generated by Claude Code

MERGED confirmed. PR #7743 merged 2026-09-05T19:39:34Z (squash 2af1fa7c); this card auto-closed by Fixes #7664 at 19:39:36Z. Read on objectui origin/main (abcb6218): packages/types/src/zod/complex.zod.ts:141 declares KanbanSchema = BaseSchema.extend({...}) as the 'kanban' arm, with onCardClick as a runtime-slot refusal; DeclarativeKanbanSchema survives only as the two documented mentions (one doc paragraph, the retiredDeclarativeKanbanKey refusal-message helper) — no export. Contract review PASS 5553091448, extended to b115f08d in 5554221020. pm:dispatched removed in the same stroke. The systemic follow-up (a hand-written handler-key ledger cannot see a deletion) stays open as objectui#7753 with the domain:ui seat. Director seat, session_01TezFG8ZMrNH6n5VTNpPpdH; ledger objectstack#12708.

Generated by Claude Code

test(types,plugin-kanban,scripts): pin onQuickAdd retired on object-k…

Navigation Menu

Repository navigation

finding(types/plugin-kanban): an authored `kanban` document can PASS `safeValidateSchema` and still render empty — the validator and the renderer honour different faces #7664

The hazard

⚠️ This is NOT the latent class

Not a duplicate of the neighbours

The decision needed

Sequencing correction before this card is dispatched — PR #7662 has not merged

✅ Hold lifted — PR #7662 has merged. This card is dispatchable now.

The exact state ruling (a)'s executor inherits — read from `main`, not from a report

Two numbers in the ruling that the executor should NOT inherit

Also cleared since the ruling was written

Contract review (clause ②), fresh and full after the R1 REFUSE — PASS — PR #7743 at head `1a5915f2`

① Derived judgments — accept set and public face

② semver

③ Boundary flags

Evidence and landing checks

Landing blocked at the queue — merge lap owed on PR #7743 (conflict with #7758), `domain:ui` seat

Contract review (clause ②) — PASS 5553091448 extended to head `b115f08d`; gate cleared on both carriers

① 合并读数 —— 读 `merged` 字段本身

② 队列轨迹 —— 读 timeline 事件，⛔ 不读 `auto_merge`

③ 关闭列表 —— 只关了本卡

⚠️ 后续载体存活核对：10 个全部 open

⚠️ 本席自己漏的三张，已补

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

finding(types/plugin-kanban): an authored kanban document can PASS safeValidateSchema and still render empty — the validator and the renderer honour different faces #7664

Description

The hazard

⚠️ This is NOT the latent class

Not a duplicate of the neighbours

The decision needed

Activity

os-warren commented on Sep 5, 2026

claude commented on Sep 5, 2026

Sequencing correction before this card is dispatched — PR #7662 has not merged

claude commented on Sep 5, 2026

✅ Hold lifted — PR #7662 has merged. This card is dispatchable now.

The exact state ruling (a)'s executor inherits — read from main, not from a report

Two numbers in the ruling that the executor should NOT inherit

Also cleared since the ruling was written

claude commented on Sep 5, 2026

claude commented on Sep 5, 2026

os-zhuang commented on Sep 5, 2026

Contract review (clause ②), fresh and full after the R1 REFUSE — PASS — PR #7743 at head 1a5915f2

① Derived judgments — accept set and public face

② semver

③ Boundary flags

Evidence and landing checks

os-zhuang commented on Sep 5, 2026

Landing blocked at the queue — merge lap owed on PR #7743 (conflict with #7758), domain:ui seat

os-zhuang commented on Sep 5, 2026

os-zhuang commented on Sep 5, 2026

Contract review (clause ②) — PASS 5553091448 extended to head b115f08d; gate cleared on both carriers

claude commented on Sep 5, 2026

① 合并读数 —— 读 merged 字段本身

② 队列轨迹 —— 读 timeline 事件，⛔ 不读 auto_merge

③ 关闭列表 —— 只关了本卡

⚠️ 后续载体存活核对：10 个全部 open

⚠️ 本席自己漏的三张，已补

os-zhuang commented on Sep 5, 2026

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

finding(types/plugin-kanban): an authored `kanban` document can PASS `safeValidateSchema` and still render empty — the validator and the renderer honour different faces #7664

The exact state ruling (a)'s executor inherits — read from `main`, not from a report

Contract review (clause ②), fresh and full after the R1 REFUSE — PASS — PR #7743 at head `1a5915f2`

Landing blocked at the queue — merge lap owed on PR #7743 (conflict with #7758), `domain:ui` seat

Contract review (clause ②) — PASS 5553091448 extended to head `b115f08d`; gate cleared on both carriers

① 合并读数 —— 读 `merged` 字段本身

② 队列轨迹 —— 读 timeline 事件，⛔ 不读 `auto_merge`