fix(spec): JoinedReportBlock, a ViewItem config and the overlay viewKind carry the shapes their doors accept (#19920)

[objectstack-fleet]

Part of #19920
Clause-②: yes (narrowing)

This PR takes the four remainder items that seat 1's release on #19920 (comment 5858219255) lists, plus remainder 5 (the flattened list overlay's type and columns), on main after PR #20286. #19920 remains open because this round's census and probes found three more sites of the same family, outside this PR's surface. See "What stays on #19920" below.

What changed

Only types change. The runtime accept set does not move: no schema's parse, no value, and no removed or renamed export. The FROM column was probed on the base c5dcb3ba0, both against the source (compiler API, tsconfig.test.json options) and against the built dist.

item	FROM	TO
1. JoinedReportBlock (ui/report.zod.ts)	unknown, because JoinedReportBlockSchema was annotated z.ZodTypeAny. Each blocks[] element of Report / ReportParsed (and so of defineReport's parameter) was unknown for the same reason.	The schema's own input type: the annotation is removed. blocks[] elements are blocks.
2. A ViewItem's config (ui/view.zod.ts)	unknown on both arms of ViewItem and ViewItemWire, because viewItemArmShape(viewKind, config) typed config as z.ZodTypeAny. The same unknown reached defineViewItem and the viewItem member of ViewMetadata, ViewMetadataParsed, AssembledViewArtifact and AssembledViewArtifactParsed.	The arm's config type: a ListView config under viewKind: 'list', a FormView config under viewKind: 'form'. config is a generic parameter now, beside viewKind.
3. The flattened overlay members' viewKind	The union of 'list' and 'form' on both members, because flattenedViewOverlayFields(kind) was not generic, so z.enum([kind]) widened. { type: 'grid', columns: ['name'], object, viewKind: 'form' } type-checked, through the list member, as all four union types above, while both doors refuse it.	'list' on the list overlay member, 'form' on the form overlay member: the function is generic in the kind.
4. The list overlay's type default	Not named anywhere.	One TSDoc paragraph on ViewMetadataParsed, and the same on AssembledViewArtifactParsed. It says, from the code, that the member declares type without .default('grid') and re-applies it in .overwrite(applyListOverlayTypeDefault). An .overwrite() returns the member's own output type, so type stays optional in the output type while every parse through that member returns it ('grid' when the body named none). On that member type is typed as the list shape's type enum.
5. The flattened list overlay's type and columns (ui/view.zod.ts)	unknown on the list overlay member of ViewMetadata, ViewMetadataParsed, AssembledViewArtifact and AssembledViewArtifactParsed, on input and on output, because listOverlayPatchFields() read both keys off ListViewShapeSchema through a cast of its shape to a record of z.ZodTypeAny. { object, viewKind: 'list', columns: 42 } type-checked as all four while that member refuses it.	The list shape's own types, both optional: type the list view type enum, columns the field list. The shape is read as typed; the schemas are the same objects, so no parse moves.

Changes beyond the five items, each forced, with what forced it

1. Declaration size (item 2). With only the generic parameter, view.zod.d.ts grows by +170,655 B (500,535 to 671,190, +34%), with 0 TS7056. The cause is that ViewItemSchema, ViewItemWireSchema and VIEW_METADATA_MEMBERS.viewItem each spell both config types out in full, at 58,764 B, 59,326 B and +60,108 B.
   • What this PR does instead. Both schemas carry an explicit type annotation through two non-exported aliases read off viewItemArmShape itself. ViewItemArmShape is the ReturnType of typeof viewItemArmShape, instantiated. ViewItemWireArmShape is that plus viewItemWireFields(), written as one mapped type. The viewItem entry of VIEW_METADATA_MEMBERS is spelled ViewItemWireSchema as typeof ViewItemWireSchema, an assertion to the schema's own type, so the emitter names the schema instead of copying it.
   • Result. view.zod.d.ts is 496,008 B, 4,527 B smaller than on the base.
   • Identity proof. Measured with an Equal probe (the mutual-conditional form) against in-memory inferred twins of both schemas. It holds for each schema, for its z.input and z.output, and for the member. A control pair answers false.
   • A spelling I measured and did not take. A first spelling of the wire annotation used an intersection. It was assignable but not identical (Equal false), so it was replaced by the mapped form.
   • The deviation. No TS error forced this assertion. Declaration size forced it. So it deviates from the dispatch's mechanism assumption 2, which allowed assertions only where a TS error forces them.
2. ADR-0122 (items 1 and 2). Three isomorphism pins in type-alias-convention.pin.test.ts turned into TS2344 ×3: Iso_ui_report__JoinedReportBlockSchema, Iso_ui_view__ViewItemSchema and Iso_ui_view__ViewItemWireSchema. They had held only because input and infer were the same erased unknown; typed, each schema carries defaults. The pins are deleted. Per check:spec-parsed-alias rule 2, three parsed-state aliases are added: JoinedReportBlockParsed, ViewItemParsed and ViewItemWireParsed. These are three new type exports on the ui entry; api-surface/ui.json and export-origins/ui.json were regenerated. The pin count is restated from 786 to 783 in the file's two prose places and its count history.
3. test-typecheck-debt.json. The report.test.ts entry shrinks by two signatures (TS18046 "'b' is of type 'unknown'" and TS2571), because blocks[] is typed now. It was regenerated with gen:test-typecheck-debt.
4. A pending release note corrected on purpose. See the next section.

Confirmation needed: a pending release note is corrected on purpose (Check Changeset stays red)

• The note. .changeset/19920-exported-types-not-unknown.md is PR fix(spec): InlineAction, ViewMetadataParsed and AssembledViewArtifact(Parsed) name their shapes, not unknown (#19920) #20260's pending entry for this card. It ended "JoinedReportBlock is not changed by this change, and still resolves to unknown."
• What this PR changes under it. Item 1 makes the second half of that sentence false in any release that carries both entries.
• The rewrite. The sentence now reads "JoinedReportBlock is not changed by this change. It stops resolving to unknown in its own entry ([finding] four more exported spec types resolve to unknown while their TSDoc promises a shape — ViewMetadataParsed, InlineAction, AssembledViewArtifact, JoinedReportBlock (the #19871 class, other sites) #19920)." That holds whichever release carries either entry. Nothing else in the note moves.
• What the gate says. node scripts/check-empty-changeset.mjs --base origin/main exits 1, naming exactly this file, in its DELIBERATE CORRECTION class. Following the gate's prescription, the file is not restored from the base, which would put the false sentence back.
• What happens next. It stays red until someone confirms the correction here. If a release consumes that entry before this PR lands, the merge keeps main's deletion and the correction is moot.

The new entry is .changeset/19920-exported-types-remainder.md: @objectstack/spec minor, Clause-②: yes (narrowing) (the diff narrows published types and also adds three exported type names, per the at-tier record 5863387565), a BREAKING banner naming every type that narrows, FROM and TO per item (four bullets, remainder 5 included), a banner and ADR-0087 reason saying that no EXISTING export changes, and the ADR-0087 marker not-required (no-migration-prescription).

Measurements

• TS7056: 0 in every spec build of this round: origin/main 5049a3cf, the pre-remainder-5 head 406335af and the head 777b5097.
• Remainder 5's cost (406335af to 777b5097): the view.zod chunk grows 495,664 to 498,393 B (+2,729, the same for .d.mts); the page.zod chunk moves -3 B; every other declaration file is byte-identical; all 128 files +5,452 B.
• The whole PR (5049a3cf to 777b5097):

module	base	head	delta
view.zod chunk (.d.ts, and the same for .d.mts)	500,191	498,393	-1,798
page.zod chunk (report and assembled views)	306,012	311,003	+4,991
index.d.ts	3,277,798	3,283,480	+5,682
system/index.d.ts	3,644,976	3,647,050	+2,074
ui/index.d.ts	524,900	524,978	+78
all 128 .d.ts / .d.mts files	30,888,091	30,910,145	+22,054 (+0.071%)

• Item 1's in-memory emit figures (+3,542 B of report.zod) are the previous run's and were not re-measured; the chunk table above supersedes them.

Reverse verification (on disk, from committed state, via scripts/ablation-replace.mjs)

All five legs re-run at 777b5097, from committed state, through scripts/ablation-replace.mjs in wrap mode; each pin file compiled under tsconfig.test.json's options:

leg	reverted to	pin file	result
item 1	JoinedReportBlockSchema: z.ZodTypeAny	joined-report-block-type.test.ts	7 x TS2578, 2 x TS18046
item 2	config: z.ZodTypeAny in viewItemArmShape	view-item-config-type.test.ts	8 x TS2578, 1 x TS18046
item 3	non-generic flattenedViewOverlayFields	view-overlay-viewkind-type.test.ts	11 x TS2578
item 4	the list overlay's type keeping the list default	view-overlay-viewkind-type.test.ts	1 x TS2322
remainder 5	the cast of the shape to a record of z.ZodTypeAny	view-overlay-viewkind-type.test.ts	6 x TS2578, 4 x TS2322

Every leg: the tool reports the mutation landed and the restore proven (blob after restore equals the HEAD blob, git diff HEAD empty), and a second check re-reads git hash-object. Control, no mutation: the three pin files compile with 0 diagnostics; git status --porcelain empty.

Tests and gates, at 777b5097

At 777b5097: spec build exit 0 (TS7056 x0); spec typecheck exit 0 (check:test-typecheck OK, 53 files / 253 errors / 140 signatures); spec check:generated exit 0; spec vitest local 560 files, 16,493 passed, 1 todo; spec test:repo 34 files, 620 passed; consumer typechecks exit 0 for lint, metadata, metadata-protocol and objectql after building their 12-package closure (rest NOT MEASURED locally, 26-package closure). dispatch-gates derived 86: 83 exit 0, check-empty-changeset exit 1 (the deliberate correction), 2 NOT MEASURED (check:dual-build-cjs-loads, check:type-check-debt, exit 3); --ran: 86 derived, 84 run, 2 NOT-MEASURED, 0 UNRUN. Lint narrowed and proven: 7 changed .ts files, 0 errors, 0 warnings; the 5 other changed files are outside eslint's configuration; no type-aware linting (eslint.config.mjs:327).

Consumer census

• This repo. Outside packages/spec, the code that imports any narrowed name is:
  • defineReport in the two example report files above;
  • Report in qa/downstream-contract/src/pipeline.report.ts;
  • ReportSchema in one platform-objects test and one downstream-contract test;
  • ViewMetadataSchema in two metadata-protocol tests;
  • AssembledViewArtifactSchema in objectql's engine.ts and one objectql test (its own static type does not move).
  • Nothing outside spec names viewItemArmShape or flattenedViewOverlayFields, which are not exported. All consumer typechecks above are green.
• objectui at the pin f8a9d0fb: one break, by design, for the seat to carry. packages/types/src/__tests__/report-chart-query-spec-parity.test.ts:725 is the inverted tripwire true satisfies IsUnknown of the spec's JoinedReportBlock. Compiled with that file's own two type definitions against this head's dist, it gives TS1360 ("Type 'true' does not satisfy the expected type 'false'"); line 726 (IsAny) still holds.
  • Its docblock says this failure is the instruction to re-run the triage and burn the local interface down, together with the ledger entry at scripts/check-spec-symbol-derivation.mjs:636.
  • It reds objectui's own type-check (tsc -p tsconfig.test.json) on its next spec bump. It does not red this repo's Console Pin Gate, which builds objectui without __tests__.
• objectui's other hits. Five other objectui files reference ViewItemSchema / ViewMetadataSchema in code; every other census hit is a comment or objectui's own local JoinedReportBlock interface. From reading them, not from compiling them: they read .success / .error of a safeParse, or cast the schema as unknown as a local interface. None reads a typed config off a parse result.
• Remainder 5 at objectui's pin f8a9d0fb. Six calls in four test files parse a body with ViewMetadataSchema, and none reads .data off the verdict, so the now-typed list-overlay type / columns reach no objectui read (from reading, not compiling).

What stays on #19920

A compiler-API census of the 2,337 non-generic exported aliases of packages/spec/src, excluding tests, with a control that must read lit: 6 aliases resolve to unknown before and after this PR, and none belongs to this family (FlowValueSlot, AssignmentValue and their Parsed, value slots; GetPublishedMetaItemResponse and its Parsed, opaque by ruling). At top-level-key level it counts 210 keys typed unknown before remainder 5 and 202 after; the 8 that left are remainder 5. Three sites of the family remain, outside this PR's surface:

• ApiError.code (input): ErrorCode is cast to a one-argument z.ZodType, whose input defaults to unknown, and makeApiErrorSchema repeats the cast. { code: 42, message: 'x' } compiles as ApiError while ApiErrorSchema refuses it at code.
• ViewFilterRule.operator (input): a z.preprocess, whose input is unknown (the cause InlineAction had). { field: 'status', operator: 42 } compiles while the door refuses it.
• The list overlay's options bag: listViewKindBlocks() returns a record of z.ZodTypeAny, so the bag is typed as a string-keyed record of unknown. options: { foo: 1, kanban: 42 } compiles as ViewMetadata while the door refuses it. The census does not scan nested keys; a probe found this one.

Acceptance notes

• The pin count holds at 783 (check:spec-parsed-alias: 1,442 bare z.input aliases, 783 pinned isomorphic, 659 paired).
• objectui tripwire, carried by the seat. packages/types/src/__tests__/report-chart-query-spec-parity.test.ts at objectui pin f8a9d0fb asserts that JoinedReportBlock is unknown (line 725). It goes red at the next objectui spec-pin bump, and its docblock says that failure is the instruction to re-run the triage and burn down the local interface. Carrier: the next objectui spec-pin bump. It cannot red this repo: the Console Pin Gate builds objectui without its __tests__.
• Fence held. FormViewSchema.layout (spec(ui): layout: 'inline' | 'grid' on object-form / FormView parse green and render as vertical — enforce-or-remove (ADR-0049) #20221), ListViewShapeSchema.tabs and the view container's body name (spec(ui): retire list.tabs and the view container's body name (2 keys); listViews + ViewTabBar and the row name already deliver both #20301) are untouched. origin/main 5049a3cf was merged once, at 1161545d, with no conflict; commits that landed on main later touch none of this PR's files.

---

Generated by Claude Code

[github-actions]

📓 Docs Drift Check

This PR changes 1 package(s): @objectstack/spec, touching 14 documentable anchor(s). ⚠️ 4 changed file(s) yielded no anchor (packages/spec/api-surface/ui.json, packages/spec/export-origins/ui.json, packages/spec/src/ui/assembled-views.zod.ts, …), so the pages documenting them are NOT COVERED by this run — this is not a clean bill of health for those files.

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

• content/docs/releases/v12.mdx (via viewKind (literal, a string literal on a changed line))

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

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

Coarse fallback — 136 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 5049a3cf4e3dd591cf5fc23856cc3a42234c9267 → packageMentionDocs.

Which tree this was computed on

This run read content/docs from f5e4830d4d9e18648da12ea273f94e0df1b6e1e4 — the merge of head 777b5097834d92c9e9565878406aa2409c597811 into base 5049a3cf4e3dd591cf5fc23856cc3a42234c9267, 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 f5e4830d4d9e18648da12ea273f94e0df1b6e1e4 && git checkout f5e4830d4d9e18648da12ea273f94e0df1b6e1e4
# afterwards, rebuild it from the two parents, which stay fetchable
git fetch origin 5049a3cf4e3dd591cf5fc23856cc3a42234c9267 777b5097834d92c9e9565878406aa2409c597811 && git checkout -B drift-repro 5049a3cf4e3dd591cf5fc23856cc3a42234c9267 && git merge --no-ff 777b5097834d92c9e9565878406aa2409c597811

node scripts/docs-audit/affected-docs.mjs --json 5049a3cf4e3dd591cf5fc23856cc3a42234c9267

⚠️ 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 5049a3cf4e3dd591cf5fc23856cc3a42234c9267 → pass the list as
args.docs, on the commit named under Which tree this was computed on.

[objectstack-fleet]

Contract review

Served-tier: CONTRACT_REVIEW_TIER
Head-sha: 35094e87739f152baa3b46c8a3250259c033208c
Local-runs: none

Inputs: card #19920 (body and all 11 comments), PR #20369 (body, 12-file list, net diff against merge-base 67047171 — identical to the local three-dot diff, 12 files, +440/−16), and the 35 check-runs on the head. Read-only throughout: git fetch plus git diff / git show against refs in the checkout; nothing built, run or re-run.

① Derived judgments

Accept set: none moves — RIGHT. The source edits are type-level only: JoinedReportBlockSchema loses its z.ZodTypeAny annotation (expression unchanged); viewItemArmShape and flattenedViewOverlayFields gain a generic parameter (bodies unchanged); ViewItemSchema / ViewItemWireSchema gain explicit type annotations (value expressions unchanged); VIEW_METADATA_MEMBERS.viewItem gains an assertion to its own type. No parse, default, refinement, key, tombstone or export is removed or renamed (api-surface/ui.json: +3 / −0). Test Core ×6, Dogfood ×3, Temporal Conformance and Spec property liveness are green on the head.

Public surface, each change named:

• JoinedReportBlock: FROM unknown TO the input type of JoinedReportBlockSchema — narrowing, RIGHT (card site 4). Carried to every blocks[] element of Report / ReportParsed (report.zod.ts:486, z.array(JoinedReportBlockSchema)) and to defineReport's parameter (:596) — RIGHT; both named in the banner. Pinned by 7 TS2578-style directives in joined-report-block-type.test.ts.
• JoinedReportBlockSchema's declared type: FROM z.ZodTypeAny TO inferred — a published const's type narrows, RIGHT, named in the changeset.
• ViewItem / ViewItemWire config: FROM unknown TO the arm's own config (ListView on list, FormView on form) — narrowing, RIGHT (remainder item 2). Carried to defineViewItem's parameter and return (view.zod.ts:5113) and to the viewItem member of ViewMetadata, ViewMetadataParsed, AssembledViewArtifact, AssembledViewArtifactParsed — RIGHT, all named. Pinned by 8 directives in view-item-config-type.test.ts.
• ViewItemSchema / ViewItemWireSchema declared types: FROM inferred TO an explicit discriminated-union annotation over two non-exported arm aliases — RIGHT as a surface statement, named. The "identical to the inferred type" claim is the dev's in-memory Equal probe and is not gate-verified; what the gates prove is that the annotation compiles on every Type Check lane and the config pins hold, so the narrowing is real and the annotation is now the contract.
• Flattened overlay members' viewKind: FROM the list / form union on both members TO each member's own literal — narrowing on the four union types, RIGHT (remainder item 3), named; pinned by view-overlay-viewkind-type.test.ts.
• NEW exports JoinedReportBlockParsed, ViewItemParsed, ViewItemWireParsed on the ui entry (api-surface/ui.json and export-origins/ui.json regenerated, +3 each): additive growth of the public surface. Adding them is RIGHT — check:spec-parsed-alias rule 2 ("the parsed state is named, or the schema is pinned isomorphic") forces a named XParsed once the three isomorphism pins cannot hold (input and infer differ by defaults once typed); the pins are deleted and the count restated 786→783 in both prose places and the history. Named in the changeset. Its DECLARATION is wrong — ②.
• VIEW_METADATA_MEMBERS.viewItem type: unchanged (typeof ViewItemWireSchema) — RIGHT.
• TSDoc on ViewMetadataParsed / AssembledViewArtifactParsed (remainder item 4): the .overwrite() type default and the list overlay's type / columns unknown — doc-only, RIGHT, pinned by two compile-time lines that go red the day that site is typed.
• test-typecheck-debt.json: report.test.ts loses TS18046 and TS2571 — consistent with blocks[] now typed; Type Check · debt ledger green. RIGHT.
• The one new site found (listOverlay type / columns unknown) stays on the card and the PR stays Part of — RIGHT for a family card; the Part-of guard is green.

Nothing in the derivations is wrong or missing.

② Semver level

• Level minor on @objectstack/spec — RIGHT. A published-type narrowing ships minor in the launch window (the fix(spec): ViewMetadata names a view body (the union of its members’ input types), not unknown #19919 / fix(spec): InlineAction, ViewMetadataParsed and AssembledViewArtifact(Parsed) name their shapes, not unknown (#19920) #20260 shape), and a new exported symbol takes at least minor; both roads meet here. Not patch, not skip-changeset, no major. The BREAKING banner names every narrowed carrier FROM→TO with the "if your code stops compiling" line; the ADR-0087 marker not-required (no-migration-prescription) is present and right (nothing objectstack migrate meta can reach). Arm (narrowing) — RIGHT.
• Clause-②: value — WRONG. PR body: Clause-②: no; changeset: Clause-②: no (narrowing). The fleet's one reader (scripts/pm/clause2-line.mjs) says the value answers whether the diff widens an accept set OR expands the public surface. This diff adds three exported names to the ui entry and had to regenerate the public-surface ledger to admit them. The fleet's written rule for exactly this case is .changeset/17779-dashboard-metric-family-single-measure.md: yes (narrowing) because "the published surface GAINS one exported symbol … a new exported symbol is the mechanical floor for in-seat contract review"; every no (narrowing) entry that explains itself (17499, 19514, 19751, 19757, 19889, 19977, 20080, 20116) does so with "no exported symbol moves", and 17499 names "check:api-surface clean with no regeneration" as the test — which this PR does not meet. The family precedents fix(spec): ViewMetadata names a view body (the union of its members’ input types), not unknown #19919 and fix(spec): InlineAction, ViewMetadataParsed and AssembledViewArtifact(Parsed) name their shapes, not unknown (#19920) #20260 declared no (narrowing) and touched no api-surface file; this PR is the first in the family to add names, and the claim's Clause-②: no was copied before the diff existed (the same claim says "adds three type exports"). The truthful spelling is Clause-②: yes (narrowing) — the reader's own case for "widens one surface and narrows another; both facts are read". No CI outcome changes with the fix (the LEVEL axis under yes wants at least minor on a moved package — already minor; ADR-0087 already reads the arm), but the line ships to consumers in CHANGELOG.md and the body line is the carrier of record, so it must be true.
  • Remedy, one changeset-only commit plus one body edit: PR body line → Clause-②: yes (narrowing); changeset line → Clause-②: yes (narrowing), and tighten the banner's "no schema's parse, no export and no value changes" to "no existing export changes" (three are added). No source, test or generated file moves; the fresh head then takes a same-head record.
• The deliberate correction — RIGHT and MINIMAL, confirmed. .changeset/19920-exported-types-not-unknown.md (PR fix(spec): InlineAction, ViewMetadataParsed and AssembledViewArtifact(Parsed) name their shapes, not unknown (#19920) #20260's pending note) moves by exactly one line, +1 / −1: "…is not changed by this change, and still resolves to unknown." becomes "…is not changed by this change. It stops resolving to unknown in its own entry ([finding] four more exported spec types resolve to unknown while their TSDoc promises a shape — ViewMetadataParsed, InlineAction, AssembledViewArtifact, JoinedReportBlock (the #19871 class, other sites) #19920)." Nothing else in the note moves. The note still exists on origin/main (862b6ce8) with the original sentence, so the correction is live, not moot: a release carrying both entries would otherwise print a false sentence. The gate's own prescription for this class is "do NOT restore it — say so on the PR and get it confirmed", and the PR body does so in its own section, naming the note and what changed under it. The red's annotation names exactly that file and no other.
• Other reds: NONE. 35 check-runs on the head: 31 success, 3 skipped, 1 failure. The failure is Check Changeset — a pull_request-only job in pr-automation.yml, not among the seven required contexts on main (TypeScript Type Check, Test Core, Dogfood Regression Gate, Build Core, Temporal Conformance, Lint & Repo Gates, Governed Surface Queue Guard — all seven success on this head). The three skips are filter-driven, not red: Build Docs (docs filter: apps/docs/**, content/**, pnpm-lock.yaml), Console Pin Gate (console filter: .objectui-sha, the console scripts, ci.yml), Packed-tarball smoke (opt-in); none of those paths is in the diff.

③ Boundary flags

No os-dev-report exists (clock-out, release 5862804129); the flags are the PR body's.

1. Deviation: annotations and one assertion forced by declaration size, not by a TS error (ViewItemSchema, ViewItemWireSchema, viewItem: … as typeof ViewItemWireSchema) — ANSWERED, accepted. The card body itself says to measure the declaration-size cost before touching a ZodTypeAny cast ([finding] AssembledInstalledPackage.manifest erodes to an index-signature type in the published .d.ts, so the assembled arm absorbs the authoring arm #19324); it was measured (+170,655 B of view.zod.d.ts inferred versus −4,527 B annotated, 0 TS7056 either way), the assertion is to the schema's own type, the arm aliases are non-exported and read off viewItemArmShape so they cannot drift, and the claim's condition — each such change named — is met in the PR body. Within the fence.
2. Three new exports and three deleted isomorphism pins — ANSWERED in ①: gate-forced (check:spec-parsed-alias rule 2), named, ledgers regenerated. What is owed is the declaration value (②).
3. Confirmation of the pending-note correction — ANSWERED in ②: right, minimal, confirmed by this record's reading.
4. Part of versus a closing keyword (the new listOverlay type / columns site) — ESCALATED to the seat, non-blocking. Part of is right as the PR stands: the site is the same family as item 2, seat 1's release bundled family remainders on this card on purpose, and the Part-of guard is green. The seat either keeps it on [finding] four more exported spec types resolve to unknown while their TSDoc promises a shape — ViewMetadataParsed, InlineAction, AssembledViewArtifact, JoinedReportBlock (the #19871 class, other sites) #19920 as remainder 5 with the measured fix (read ListViewShapeSchema.shape without the cast, +3,154 B, 0 TS7056) or files it; either way the item-4 TSDoc and the two pin lines in view-overlay-viewkind-type.test.ts are corrected with it, by design.
5. objectui tripwire at pin f8a9d0fb (packages/types/src/__tests__/report-chart-query-spec-parity.test.ts:725, true satisfies IsUnknown of JoinedReportBlock, TS1360 against this head's dist) — ESCALATED to the seat, non-blocking for this repo. Verified structurally: the Console Pin Gate runs turbo run build / pnpm run build only, never objectui's tsconfig.test.json, so it cannot red here — and it was skipped on this head by the console filter, so it is NOT MEASURED. AGENTS.md's ship-together rule (Post-Task Add Changesets and GitHub Actions automation #4) covers removals and renames the sibling imports; this is a narrowing whose sibling break is a designed tripwire whose docblock says the failure is the instruction. The seat carries an objectui follow-up (burn down the local JoinedReportBlock interface; ledger at scripts/check-spec-symbol-derivation.mjs:636) with or before the next pin bump.
6. Concurrency — origin/main moved once on this diff's files since the merge-base: PR feat(spec)!: form layout accepts only vertical | horizontal — the inline and grid arms retired (#20221) #20262 (spec(ui): layout: 'inline' | 'grid' on object-form / FormView parse green and render as vertical — enforce-or-remove (ADR-0049) #20221) landed FormViewSchema.layout at view.zod.ts 4038–4121, disjoint from every hunk here (4895+, 5027+, 5078+, 5292+, 5923+, 6713+, 6750+); none of the other 11 files moved. Open PRs feat(spec)!: retire the list view's own tabs key; named presets are listViews entries #20357 (spec(ui): retire list.tabs and the view container's body name (2 keys); listViews + ViewTabBar and the row name already deliver both #20301) and feat(spec)!: retire the connector resilience family — health (probe + breaker), status and nested webhooks, sixteen keys nothing read (#20273) #20350 (spec(integration): retire the connector health-probe, circuit-breaker, authored status and nested webhooks keys (16), which nothing enforces #20273) share files: ordinary concurrency, the rework head merges main once (the pin count and test-typecheck-debt.json are the collision-prone files if feat(spec)!: retire the connector resilience family — health (probe + breaker), status and nested webhooks, sixteen keys nothing read (#20273) #20350 lands first). Fence held: FormViewSchema.layout, ListViewShapeSchema.tabs and the container body name are untouched.
7. NOT MEASURED locally by the dev (check:dual-build-cjs-loads, check:type-check-debt, the example-app and workspace typechecks) — answered by CI on this head: Build Core, Type Check · debt ledger, Type Check · workspace, Type Check · consumer gates and Type Check · source gates are all green.

Implemented-by: claude/issue-19920-exported-types-remainder
Reviewed-by: session_01ARcDurZ5j34RdqsGgc4jgH

VERDICT: FAIL

---

Generated by Claude Code

[objectstack-fleet]

Contract review

Served-tier: CONTRACT_REVIEW_TIER
Head-sha: 777b5097834d92c9e9565878406aa2409c597811
Local-runs: none

Inputs: card #19920 (body and all 13 comments), PR #20369 (body, 12-file list, net diff against merge base 5049a3cf; GitHub's diff and the local git diff 5049a3cf..777b5097 agree at 12 files, +482/−20, and the merge base against today's main 15bf186f is still 5049a3cf), and the 42 check-runs on the head. Read-only throughout: git fetch plus git show / git diff / git grep against refs in the checkout; nothing built, run or re-run.

① Derived judgments

Accept set: none moves — RIGHT. Every source edit is at the type level and leaves the value expression as it was:

• report.zod.ts: JoinedReportBlockSchema drops its z.ZodTypeAny annotation; the lazySchema factory of strictObject(...) with .superRefine(checkReportOrder) is unchanged. The schema is not recursive and no probe in recursive-schema-input-assertions.ts names it, so nothing pinned the annotation.
• view.zod.ts: viewItemArmShape and flattenedViewOverlayFields gain a generic parameter with bodies unchanged; flattenedViewOverlayFields is called only with the literals 'list' (:5901) and 'form' (:5930). listOverlayPatchFields() reads ListViewShapeSchema.shape as typed instead of through a cast to a record of z.ZodTypeAny: the same shape.columns.optional() and shape.type.unwrap().optional().meta({ default }) on the same objects (type carries .default('grid') and columns is declared on that shape), so the served JSON Schema and the parse are byte-identical. ViewItemSchema / ViewItemWireSchema gain explicit annotations; VIEW_METADATA_MEMBERS.viewItem gains an assertion to its own type. No parse, default, refinement, key, tombstone or export is removed or renamed. Test Core ×6, Dogfood ×3, Temporal Conformance, Spec property liveness and Build Core are green on the head.

Public surface, each change named:

• JoinedReportBlock: FROM unknown TO the input type of JoinedReportBlockSchema — narrowing, RIGHT (card site 4, the last of the card's four). Carried to every blocks[] element of Report / ReportParsed (z.array(JoinedReportBlockSchema) at :486) and defineReport's parameter (:596) — RIGHT, all named in the banner. Pinned by joined-report-block-type.test.ts.
• JoinedReportBlockSchema's declared type: FROM z.ZodTypeAny TO inferred — a published const's type narrows, RIGHT, named in the changeset.
• ViewItem / ViewItemWire config: FROM unknown TO the arm's own config (typeof ListViewSchema on list, typeof FormViewSchema on form; .describe() returns the same type) — narrowing, RIGHT (remainder 2). Carried to defineViewItem and to the viewItem member of ViewMetadata, ViewMetadataParsed, AssembledViewArtifact, AssembledViewArtifactParsed — RIGHT, named. Pinned by view-item-config-type.test.ts.
• ViewItemSchema / ViewItemWireSchema declared types: FROM inferred TO an explicit discriminated-union annotation over two non-exported arm aliases — RIGHT as a surface statement. Checked structurally: lazySchema returns its factory's own type, strictObject returns a closedObject of z.object(...).strict() (so $strict), the wire arms end in .strip() (so $strip), the arm alias is the instantiated ReturnType of viewItemArmShape itself, and the wire alias's mapped type is exact because viewItemBaseShape() and viewItemWireFields() (isPinned, sortOrder, columnState) share no key. Full identity with the inferred twin is the dev's Equal probe, not gate-verified; every Type Check lane proves the annotation compiles and the config pins hold, so the annotation is now the contract.
• Flattened overlay members' viewKind: FROM 'list' | 'form' on both members TO each member's own literal — narrowing on the four union types, RIGHT (remainder 3), named; pinned by view-overlay-viewkind-type.test.ts.
• Flattened list overlay's type / columns (remainder 5): FROM unknown TO the list shape's own optional types on the list overlay member of the four union types, input and output — narrowing, RIGHT, named as the changeset's fourth bullet; pinned by the IsUnknown tuple, two member probes and four union probes in view-overlay-viewkind-type.test.ts.
• NEW exports JoinedReportBlockParsed, ViewItemParsed, ViewItemWireParsed on the ui entry — additive growth, RIGHT and forced: typed, each schema carries a default (type with .default('tabular') on the block; the list config's type default on the arm), so input and infer differ and check:spec-parsed-alias rule 2 wants the named Parsed sibling. The three isomorphism pins are deleted with the count restated 786→783 in both prose places and the history (783 pinned + 659 paired = 1,442 bare, consistent). api-surface/ui.json and export-origins/ui.json regenerated (+3 each); no other api-surface ledger lists these names (root.json does not carry JoinedReportBlock), so nothing else owes regeneration; check:generated is green in Lint & Repo Gates.
• VIEW_METADATA_MEMBERS.viewItem type: unchanged (typeof ViewItemWireSchema) — RIGHT.
• TSDoc on ViewMetadataParsed / AssembledViewArtifactParsed (remainder 4): the .overwrite(applyListOverlayTypeDefault) default, absent from the output type, type typed as the list shape's enum — doc-only, RIGHT; applyListOverlayTypeDefault is typed T → T, so the sentence is true of the code; pinned by IsOptionalKey in the overlay pin file. The earlier sentence claiming type / columns unknown on that member is gone from both source files and from the PR body.
• test-typecheck-debt.json: report.test.ts loses TS18046 and TS2571 — consistent with blocks[] now typed; Type Check · debt ledger green. RIGHT.

Nothing derived is wrong or missing.

② Semver level

• Level minor on @objectstack/spec — RIGHT. A new exported symbol takes at least minor; a published-type narrowing ships minor in the launch window (the fix(spec): ViewMetadata names a view body (the union of its members’ input types), not unknown #19919 / fix(spec): InlineAction, ViewMetadataParsed and AssembledViewArtifact(Parsed) name their shapes, not unknown (#19920) #20260 shape). Not patch, not skip-changeset, no major.
• Clause-②: yes (narrowing) — RIGHT, in both carriers. The changeset's line 7 and the PR body's line 2 both read exactly Clause-②: yes (narrowing); no Clause-②: no survives anywhere in the body. This is the reader's own case (scripts/pm/clause2-line.mjs: a diff that widens one surface and narrows another, both facts read) and the fleet's precedent for the same shape (.changeset/17779-dashboard-metric-family-single-measure.md, on main). REWORK item 1 holds: the BREAKING banner now says no existing export changes and three parsed-state type names are added, nothing removed or renamed; the ADR-0087 reason says the same; the marker token is unchanged, not-required (no-migration-prescription), and right (nothing objectstack migrate meta can reach). The banner names every narrowed carrier (JoinedReportBlock, Report, ReportParsed, ViewItem, ViewItemWire, the four union types, defineReport / defineViewItem) with FROM → TO per item and the "if your code stops compiling" line. Lint & Repo Gates, which runs check-adr-0087-registration and check-changeset-no-major, is green.
• The DELIBERATE CORRECTION — still RIGHT and MINIMAL, confirmed on this head. .changeset/19920-exported-types-not-unknown.md moves by exactly one line, +1 / −1: "…is not changed by this change, and still resolves to unknown." becomes "…is not changed by this change. It stops resolving to unknown in its own entry ([finding] four more exported spec types resolve to unknown while their TSDoc promises a shape — ViewMetadataParsed, InlineAction, AssembledViewArtifact, JoinedReportBlock (the #19871 class, other sites) #19920)." Nothing else in the note moves. It still exists on origin/main (15bf186f, blob bce741af) with the original sentence, so the correction is live, not moot, and the new sentence is true in any release that carries either entry. The red's annotations on this head (Check Changeset, job 108811698945) name exactly that file and no other, with the gate's two-class text; the class is DELIBERATE CORRECTION, not COLLISION (the PR's own entry is 19920-exported-types-remainder.md, a different filename), whose prescribed remedy is do NOT restore it, say so on the PR and get it confirmed. The PR body does so in its own section. This record confirms it.
• Other reds: NONE. 42 check-runs on the head: the only failures are the two Check Changeset runs (the same pull_request-only job across two pr-automation runs on this head). All seven required contexts on main, read from the branch rules (TypeScript Type Check, Test Core, Dogfood Regression Gate, Build Core, Temporal Conformance, Lint & Repo Gates, Governed Surface Queue Guard), are success. The skips are filter-driven, not red: Build Docs (docs filter), Console Pin Gate (console filter: .objectui-sha, the console scripts, ci.yml, none in the diff), Packed-tarball smoke (opt-in), and Auto Label / Check PR Size on the second run only.

③ Boundary flags

The dev's report is 5864550924 (open_questions: none; 7 deviations; 4 out_of_scope_findings), plus the four REWORK items and the PR body's flags.

1. REWORK item 3, the single origin/main merge — HOLDS. This round added exactly one merge, 1161545d (parents 35094e877, the FAIL-record head, and 5049a3cf); the other merge in the PR, 04f5aef5, predates 35094e877 and was already in the head the FAIL record reviewed. No conflict; the merge brought PR feat(spec)!: form layout accepts only vertical | horizontal — the inline and grid arms retired (#20221) #20262's FormViewSchema.layout region (+39/−3 in view.zod.ts between 67047171 and 5049a3cf), disjoint from every hunk here. The dev's "not merged a second time" is verified: the 10 commits on main since 5049a3cf touch none of the 12 files, and the merge base is unchanged.
2. REWORK item 2, remainder 5 — HOLDS. Fixed by the PR body's own measured route (read ListViewShapeSchema.shape as typed); re-measured at +2,729 B per view.zod chunk with 0 TS7056 (the dev's measurement; the chunk table is not gate-verified, and the whole PR leaves the view.zod chunk 1,798 B smaller than the base). The item-4 TSDoc and the pin lines were corrected with it.
3. REWORK item 4, the report and its pr_body_delta — HOLDS. Every delta row is applied in the body (line 1 Part of #19920, line 2 Clause-②: yes (narrowing), table row 5, "five items", the new Measurements, Reverse verification, Tests and gates, Consumer census, What stays, Acceptance notes); no stale "four items", no old head sha, no Clause-②: no.
4. Part of #19920 versus Fixes — ANSWERED: Part of is RIGHT as the PR stands. The REWORK's own condition for Fixes was "if no site of the card remains" after a family census; the dev's widened census and probes found three family sites (ApiError.code, ViewFilterRule.operator, the list overlay's options bag), and seat 1's review and release made [finding] four more exported spec types resolve to unknown while their TSDoc promises a shape — ViewMetadataParsed, InlineAction, AssembledViewArtifact, JoinedReportBlock (the #19871 class, other sites) #19920 the family's closing card (5857792889, 5858219255). The Part-of guard and "The card this PR closes must claim this branch" are green. The seat may instead close the card on its four named sites and file the three on a fresh family card; that is a body-line edit plus a card, not a code change, and not blocking either way.
5. Whether any of the three new sites belongs inside this PR — ANSWERED: NO, each stays on the card for the seat. None meets the adjacent-defect rule the seat applied to remainder 5 (same class, mechanical, unclaimed, same gate family), and each is outside claim 5863213367's file surface, which says stop on breach: ApiError.code lives in api/error-code-ledger.zod.ts / api/contract.zod.ts (a different entry, with in-repo consumers in metadata-protocol and types); ViewFilterRule.operator is a z.preprocess at view.zod.ts:923 whose input type is a contract choice (whether the legacy spellings the preprocess folds stay in the input type), not a mechanical edit; the options bag is listViewKindBlocks() (:5669), a runtime loop over the kinds returning a record of z.ZodTypeAny, whose typed form is a mapped type over the kind blocks with an unmeasured declaration cost. Escalated to the seat for routing only (this card or a new one), non-blocking.
6. Deviation: annotations and one assertion forced by declaration size, not by a TS error — ANSWERED, accepted, unchanged since the FAIL record: measured (+170,655 B inferred versus −4,527 B annotated, 0 TS7056 either way), the assertion is to the schema's own type, the aliases are non-exported and read off viewItemArmShape so they cannot drift, and each such change is named in the PR body as the claim requires.
7. Report deviations 1–4 (worktree from the remote branch, merge-first route order, widened census, items 1–4 ablations re-run) — process notes with no contract effect; the history shows no rewrite (35094e877 remains an ancestor of the head).
8. NOT MEASURED locally by the dev (check:dual-build-cjs-loads, check:type-check-debt, the @objectstack/rest typecheck) — answered by CI on this head: Build Core, Type Check · debt ledger, Type Check · workspace, Type Check · consumer gates and Type Check · source gates are all green.
9. objectui tripwire at pin f8a9d0fb (report-chart-query-spec-parity.test.ts:725, true satisfies IsUnknown of JoinedReportBlock) — ESCALATED to the seat, non-blocking, unchanged: a designed tripwire whose docblock says the failure is the instruction; the Console Pin Gate builds objectui without __tests__ and was skipped on this head by the console filter, so it is NOT MEASURED here. Carrier: the next objectui spec-pin bump, recorded in the PR's Acceptance notes.
10. Nit, non-blocking: the PR title names three of the five items; the changeset's title, the line that ships, names all four narrowings. No carrier reads the PR title.
11. Fence held: FormViewSchema.layout (spec(ui): layout: 'inline' | 'grid' on object-form / FormView parse green and render as vertical — enforce-or-remove (ADR-0049) #20221), ListViewShapeSchema.tabs and the container body name (spec(ui): retire list.tabs and the view container's body name (2 keys); listViews + ViewTabBar and the row name already deliver both #20301) are untouched by every hunk. Docs Drift flags content/docs/releases/v12.mdx as release-owned and read-only; it is not edited here.

Implemented-by: claude/issue-19920-exported-types-remainder
Reviewed-by: session_01ARcDurZ5j34RdqsGgc4jgH

VERDICT: PASS

---

Generated by Claude Code