fix(spec): InlineAction, ViewMetadataParsed and AssembledViewArtifact(Parsed) name their shapes, not unknown (#19920)

[objectstack-fleet]

Part of #19920

Clause-②: no

Three of the four sites #19920 names now resolve to the shape their TSDoc promises. JoinedReportBlock is the remainder: its region of report.zod.ts is held by the open PR #20238 (#20161), and the dispatch put this site after that PR merges. #19920 remains open for it. The measured options for it are under "The remainder" below.

What changed

Types only. No schema, no parse, no export and no declared type of any schema moves. Each alias was derived from a schema whose own static type erases to unknown, so any value type-checked against it. Each is now derived from the member schema the parse actually runs, the way PR #19919 re-derived ViewMetadata.

name	FROM (all unknown at e0f17a37)	TO
InlineAction (ui/action.zod.ts)	z.input of typeof InlineActionSchema. The schema is a z.preprocess, whose input type is the preprocess function's unknown parameter.	z.input of (typeof InlineActionSchema)['out']: the pipe's out member, the .pick()ed action object.
ViewMetadataParsed (ui/view.zod.ts)	z.infer of typeof ViewMetadataSchema. The union's members are cast to z.ZodTypeAny where it is built.	z.infer over (typeof VIEW_METADATA_MEMBERS)[ViewMetadataBranch]: the members' OUTPUT union, the same record ViewMetadata reads its input types from.
AssembledViewArtifact (ui/assembled-views.zod.ts)	z.input of typeof AssembledViewArtifactSchema. Same cast.	z.input over the VIEW_METADATA_MEMBERS entries minus container: the three members that schema's union is mapped from.
AssembledViewArtifactParsed	z.infer of the same schema. Same cast.	z.infer over the same three members.

• diagnoseViewMetadata: the one edit the new type forces. Its success branch returned data: parsed.data, and parsed.data is unknown for the same member cast. Without an edit that line is TS2322 (measured below). It now asserts parsed.data to ViewMetadataParsed, with a comment saying why that holds. At runtime the union's output IS the accepting member's output, since the union's .check() transforms nothing. No value changes. A new test asserts diagnosis.data deep-equals the member's own parse output for every member.
• Every changed TSDoc states what the type does NOT express. The schema is still the only judge: the preprocess folds and strips, and refinements are not types. For InlineAction, the legacy type: 'navigation' and to spellings are refused by the type while the door still folds them. That is pinned in both directions.
• The JoinedReportBlockSchema z.ZodTypeAny annotation and the member casts inside ViewMetadataSchema / AssembledViewArtifactSchema are all untouched.
• Changeset .changeset/19920-exported-types-not-unknown.md: minor on @objectstack/spec, Clause-②: no (narrowing) with the BREAKING banner (a narrowing of published TYPES; the runtime accept set does not move) and the ADR-0087 not-required (no-migration-prescription) disposition, matching the fix(spec): ViewMetadata names a view body (the union of its members’ input types), not unknown #19919 precedent; FROM and TO per name, plus the "if your code stops compiling" instruction.
• .changeset/view-metadata-type-not-unknown.md is the unreleased fix(spec): ViewMetadata names a view body (the union of its members’ input types), not unknown #19919 entry. It said "ViewMetadataParsed is not changed by this release: it is still unknown", which this branch makes false if both entries ship in one release. It now reads "ViewMetadataParsed is not changed by this change. It is re-derived from the same members, as their output types, by 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. My own entry's JoinedReportBlock sentence is worded the same way.

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

check-empty-changeset refuses this PR because it changes .changeset/view-metadata-type-not-unknown.md, a changeset it did not add. This is the gate's DELIBERATE CORRECTION class, not a filename collision:

• Note: the pending (unreleased) fix(spec): ViewMetadata names a view body (the union of its members’ input types), not unknown #19919 entry for ViewMetadata.
• What changed under it: this PR re-derives ViewMetadataParsed, so the note's sentence "ViewMetadataParsed is not changed by this release: it is still unknown" would ship false in any release that carries both entries.
• The rewrite: that one sentence now reads "ViewMetadataParsed is not changed by this change. It is re-derived from the same members, as their output types, by 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.

Per the gate's own prescription, the file is ⛔ not restored from the base, which would put the false sentence back. Check Changeset stays red until the correction is confirmed here in writing. It is not a required context. If a release consumes the #19919 entry before this PR lands, the correction becomes moot: the merge resolves by keeping main's deletion.

Measurements

All readings are TypeScript compiler-API reads of the package's own tsconfig.json / tsconfig.test.json, unless named otherwise.

1. The premise holds. At base e0f17a37: ViewMetadataParsed, InlineAction, AssembledViewArtifact, AssembledViewArtifactParsed and JoinedReportBlock all have type flag Unknown. The control ViewMetadata (re-derived in PR fix(spec): ViewMetadata names a view body (the union of its members’ input types), not unknown #19919) is NOT unknown, and InlineActionParsed was never unknown. At head, the four changed names are not unknown, and JoinedReportBlock still is.

2. Declaration cost. No cast is removed; each alias is emitted verbatim. pnpm --filter @objectstack/spec build was run on both trees in one lock turn:

   • total .d.ts bytes: 30,830,702 at e0f17a37 → 30,836,475 at 61b382d9 (+5,773, +0.019%, TSDoc and alias text);
   • affected chunks: view.zod 499,735 → 500,541, action.zod 76,621 → 77,606, page.zod 304,750 → 305,845;
   • TS7056 occurrences in the build log: 0 on both.

   The casts are real declaration-size dodges, which is why this PR derives from the members and leaves the casts alone. In-memory declaration emit, replacing each union's z.ZodTypeAny tuple with the real member tuple:

   • view.zod.d.ts: 500,881 → 663,301 bytes (+32%); ViewMetadataSchema's own declaration grows 323 → 162,743 bytes (3,995 lines);
   • assembled-views.zod.d.ts: 8,824 → 64,776 bytes (×7.3); the schema's declaration grows 306 → 56,258 bytes.

   Neither emits TS7056. The same root fix would also have removed the diagnoseViewMetadata assertion, so that assertion is the cheaper of the two ways to type data.

3. What the type change forces. Census, whole tree, git grep -w excluding .md/.mdx: outside packages/spec, nothing names the four types or diagnoseViewMetadata. Inside, only view.zod.ts itself and six test files do. An in-memory ablation removing the assertion yields exactly one diagnostic: view.zod.ts TS2322 "Type 'unknown' is not assignable to type 'ViewMetadataParsed'". The six census test files carry 8 diagnostics under tsconfig.test.json, identical on base and head apart from line numbers, all in the ledgered view.test.ts debt.

4. Consumer compile.

   • objectui at the pinned f8a9d0fb names none of the three changed types. It names JoinedReportBlock, which this PR leaves alone.
   • cloud (local checkout 48d7066) names none of the four. The control leg (defineStack) hits 18 files.
   • No consumer package in this repo imports them, so there is no consumer suite to run beyond @objectstack/spec's own.

Reverse verification

In-memory ablation: each alias reverted to its base spelling through a compiler-host override, with the anchor matched exactly once and nothing written to disk. The pin file is then compiled under tsconfig.test.json, and every @ts-expect-error pin turns red:

alias reverted	pin file	result
InlineAction	inline-action-type.test.ts	4 × TS2578 (unused directive)
ViewMetadataParsed	view-metadata-type.test.ts	3 × TS2578
AssembledViewArtifact	assembled-view-artifact-type.test.ts	3 × TS2578
AssembledViewArtifactParsed	assembled-view-artifact-type.test.ts	1 × TS2578

With the fix in place the three pin files compile with 0 diagnostics. tsc -p tsconfig.test.json --listFilesOnly lists all three, among 523 test files.

Tests

Final head f8792c93. Its code is identical to 61b382d9; cf123662 and f8792c93 touch only .changeset/. Round 2 (f8792c93, the changeset level / arm / banner / marker only) re-ran the 17 derived families that read .changeset plus check:spec-changes: check-adr-0087-registration exit 0 ([BREAKING+clause-②-narrowing] not-required (no-migration-prescription)), check-changeset-no-major exit 0 (including the level axis driven with this PR's payload), and check-empty-changeset exit 1 on the deliberate correction only. The other 66 stand at their cf123662 reading. Heavy runs went through scripts/pm/os-verify-lock.sh, and each exit code was written to disk before it was read.

• Build, typecheck and tests at 61b382d9 (lock turn 1):
  • pnpm --filter @objectstack/spec build: exit 0, TS7056 ×0;
  • pnpm --filter @objectstack/spec typecheck: exit 0, "check:test-typecheck: OK — @objectstack/spec's test layer compiles under packages/spec/tsconfig.test.json; 53 file(s) / 255 error(s) / 142 pinned signature(s) held in test-typecheck-debt.json";
  • pnpm --filter @objectstack/spec test: "Test Files 550 passed (550)", "Tests 16120 passed | 2 todo (16122)".
• Generated artifacts and pins at cf123662 (lock turn 2):
  • pnpm --filter @objectstack/spec check:generated: exit 0, "All 15 generated artifacts are up to date";
  • the three pin files by name, vitest run --project local --maxWorkers=2: "Test Files 3 passed (3)", "Tests 17 passed (17)".
• Derived gate union at cf123662: node scripts/pm/dispatch-gates.mjs --commands, 83 commands, all run.
  • 80 exit 0. Among them:
    • check:api-surface: "public API surface + factory signatures unchanged";
    • check:exported-any: "no exported type resolves to any: 2376 types + 1452 schemas";
    • check:export-origins: "5213 exports across 18 entry points resolve exactly as recorded";
    • check:docs: "226 generated files in sync";
    • check:dual-source-exports, check:entry-nameability, check:liveness, check:spec-parsed-alias, check:test-source-alias, check:issue-citations, check:nul-bytes;
    • check:lean-entry-closure and check:doc-formula-expressions, measured after building their closures.
  • check-empty-changeset --base origin/main: exit 1, the deliberate correction above, red on purpose.
  • check:dual-build-cjs-loads and check:type-check-debt: exit 3, PREREQUISITE NOT MET. NOT MEASURED: both need the whole ./packages/* build closure, which CI builds.
  • dispatch-gates.mjs --ran over the exit-coded record: "83 derived, 81 run, 2 NOT-MEASURED, 0 UNRUN".
• Lint, narrowed and proven. Repo-wide pnpm lint is CI's. I ran eslint --no-inline-config --format json over the six changed .ts files:
  • the JSON counts 6 files, 0 errors, 0 warnings;
  • the population is eslint.config.mjs's own TS/JS globs, so the two .changeset/*.md files are outside it;
  • that config "never enables type-aware linting (no parserOptions.project, no typed @typescript-eslint rules) for ANY file" (eslint.config.mjs:327), so this diff cannot move the verdict on any untouched file.
• Consumer suites: none owed. No package outside @objectstack/spec imports the four types or diagnoseViewMetadata (census above).
• NOT MEASURED, left to CI:
  • the Type Check workspace and consumer lanes;
  • Test Core shards;
  • Dogfood;
  • Build Core;
  • the two gates above.

The remainder: JoinedReportBlock

• Why it is not here. PR fix(spec): a joined report draws no chart — retire blocks[].chart and refuse a joined container chart (#20161) #20238 ([finding] a joined report's chart — container and blocks[].chart — parses and is never drawn, while the liveness ledger names the joined branch as its reader #20161, +77/−14 in report.zod.ts) edits the joined-report block schema this type is derived from, and it is still open (draft). The dispatch fixed the order: this site lands on the merged main.

• Measured fork, for whoever takes it. JoinedReportBlockSchema is annotated z.ZodTypeAny. I did an in-memory declaration emit of report.zod.ts on current main, with the annotation removed. It produced no TS7056 and no other diagnostic. report.zod.d.ts grows 18,608 → 34,390 bytes (×1.85): the block schema's declaration grows to 7,582 bytes, and ReportSchema doubles (8,677 → 16,937) because blocks: now inlines the block type. So the annotation buys declaration size, not an escape from TS7056. The two options are:

  • remove the annotation and derive the type from the schema;
  • keep the annotation and derive the type from an explicitly named shape.

  Both need re-measuring on the post-fix(spec): a joined report draws no chart — retire blocks[].chart and refuse a joined container chart (#20161) #20238 tree.

• objectui tripwire. objectui at the pin holds an inverted pin, true satisfies IsUnknown of the spec's JoinedReportBlock, in packages/types/src/__tests__/report-chart-query-spec-parity.test.ts. Its docblock says the day the spec types this, the pin stops compiling, and "the failure is the instruction: re-run the triage and burn it down". The Console Pin Gate only builds objectui (the types build config excludes __tests__/), so that pin reds objectui's own type-check on its next spec bump, not this repo's CI. Measured for the follow-up; this PR does not move it.

Acceptance notes

• check:spec-parsed-alias recognises a bare alias only in the spelling z.input of typeof the schema. Its population drops 1443 → 1441 bare aliases (paired 657 → 655), because InlineAction and AssembledViewArtifact now use member derivations, as ViewMetadata has since fix(spec): ViewMetadata names a view body (the union of its members’ input types), not unknown #19919. Both …Parsed siblings still exist; the gate just no longer sees the pairs. Noted, not filed. Carrier: none.
• Same family, nested: viewItemArmShape(viewKind, config: z.ZodTypeAny) makes config unknown on both members of ViewItem and ViewItemWire (measured). So it is unknown on the viewItem member of every union above too. Reported to the dispatching seat to fold into this family's closing card; not changed here.
• Zone 3's "a value missing a required key is a type error" pin cannot be expressed for InlineAction: every key of its input is optional (type has a default; name and label are .partial()). Its pins are unknown, the two legacy spellings, and a scalar.

---

Generated by Claude Code

[github-actions]

📓 Docs Drift Check

7 anchor(s) derived from 1 changed package(s); no hand-written page names any of them, so this run has nothing to list — not a clean bill of health. This check sees only pages that NAME a derived anchor: one that documents this change in prose, or enumerates it in an authoring dialect, names none and stays invisible to it on every run.

What this run could not see

• 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 a9fb83ef06a079a938c97d1ea5e364705e1bc13b → packageMentionDocs.

Which tree this was computed on

This run read content/docs from a83d2033447179094cf428aae5b3a556faa115a5 — the merge of head f8792c931bf34e113314e158f2edde1b3cfcbecd into base a9fb83ef06a079a938c97d1ea5e364705e1bc13b, 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 a83d2033447179094cf428aae5b3a556faa115a5 && git checkout a83d2033447179094cf428aae5b3a556faa115a5
# afterwards, rebuild it from the two parents, which stay fetchable
git fetch origin a9fb83ef06a079a938c97d1ea5e364705e1bc13b f8792c931bf34e113314e158f2edde1b3cfcbecd && git checkout -B drift-repro a9fb83ef06a079a938c97d1ea5e364705e1bc13b && git merge --no-ff f8792c931bf34e113314e158f2edde1b3cfcbecd

node scripts/docs-audit/affected-docs.mjs --json a9fb83ef06a079a938c97d1ea5e364705e1bc13b

⚠️ That checkout carried uncommitted changes, so the commit above does not fully identify what was read.

[objectstack-fleet]

Confirmed: the deliberate correction of a pending release note. domain:spec seat 1, session_01Rjy9MeetSfq34PKn81CRiN, reviewer of record for this PR, 2026-09-27T17:05Z.

What was corrected

• Note: .changeset/view-metadata-type-not-unknown.md, the unreleased fix(spec): ViewMetadata names a view body (the union of its members’ input types), not unknown #19919 entry (ViewMetadata no longer unknown).
• Old sentence: 「ViewMetadataParsed is not changed by this release: it is still unknown」.
• Why it had to change: this PR re-derives ViewMetadataParsed from the VIEW_METADATA_MEMBERS output types. If both entries ship in one release, the old sentence is false.
• New sentence: 「ViewMetadataParsed is not changed by this change. It is re-derived from the same members, as their output types, by 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.

⛔ Do not restore the base text, and ⛔ do not apply skip-changeset. This PR also adds a changeset of its own.

The red check

• Check Changeset (check-empty-changeset.mjs, the DELIBERATE CORRECTION class, ruling D on [finding] the skip-changeset label suppresses the DELIBERATE-CORRECTION refusal whose own text says 「no label and no diff shape makes that safe」 — declared contract against enforced behaviour #18375) stays red by design, per its own prescription in .github/workflows/pr-automation.yml.
• That job runs on pull_request only (no merge_group) and is not a required context, so its red blocks no merge.
• This comment is the written confirmation the gate asks for.

[objectstack-fleet]

Contract review

Served-tier: CONTRACT_REVIEW_TIER
Head-sha: f8792c931bf34e113314e158f2edde1b3cfcbecd

① Derived judgments

• The four aliases are the true shapes — RIGHT. tsc under tsconfig.test.json options over a probe file (archived at review/probes/type-probe.ts): Equal proofs hold for all four. InlineAction = z.input of the pipe's out member, with exactly the 13 picked keys; InlineActionParsed = z.output of that member = z.output of the pipe. ViewMetadataParsed = the union of z.output of the four VIEW_METADATA_MEMBERS. AssembledViewArtifact / AssembledViewArtifactParsed = the union of z.input / z.output of viewItem, listOverlay, formOverlay. Compiler-API flags at head (probes/census.out): the four are not unknown; controls ViewMetadata and InlineActionParsed not unknown; JoinedReportBlock still unknown (flags=2). z.input / z.infer of all three SCHEMAS are still unknown, so no cast moved.
• InlineAction as the pre-parse shape — RIGHT, and the fold is documented. The out member is a strict .pick(): {} type-checks and parses (type defaults to script); icon / body are refused by type and door alike (unrecognized_keys); type: 'url' with no target is admitted by the type and refused by the door (refinement), which the TSDoc names. Legacy type: 'navigation' / to are compile errors and still fold at the door to type: 'url' + target (probes/runtime-probe.out); TSDoc, changeset banner and pin all say so. The published reference table (content/docs/references/ui/action.mdx, InlineAction section) never listed to or navigation, so the type now matches the published doc.
• ViewMetadataParsed as the output union — RIGHT, one unnamed gap. The list overlay's .overwrite(applyListOverlayTypeDefault) is not in the type: listOverlay output has type optional (probe A passed) while every parse output carries type: 'grid' (a column-less patch parses to grid, runtime probe). Wider than the runtime output, so sound, but the TSDoc's "defaults applied" over-promises for this one key and neither TSDoc nor changeset names it. Non-blocking.
• Assembled pair minus container — RIGHT. AssembledViewArtifactSchema is z.union over exactly VIEW_METADATA_BRANCHES minus container (arity 3), refuses a container body, and its outputs deep-equal the member outputs (probes/door-probe.out, runtime-probe.out). Type side: a View-typed value and a ViewMetadata-typed value are not assignable to AssembledViewArtifact (TS2322), every artifact is assignable to ViewMetadata, and a container literal is TS2353.
• diagnoseViewMetadata assertion — SOUND for every branch. ViewMetadataSchema is preprocess → union .check(focusClaimedBranch()), and the check returns before touching anything when issues.length === 0, so the union's output is the first accepting member's output on the STRIPPED body. Probed one body per member plus a decorated list body (sort[].id) and a column-less patch: diagnosis.data deep-equals ViewMetadataSchema.parse(body) and VIEW_METADATA_MEMBERS[branch].parse(stripped), and branch is the first accepting member, for viewItem, container, listOverlay (three bodies) and formOverlay. In-memory ablation of the assertion: exactly 1 diagnostic, view.zod.ts TS2322 at line 6217.
• Consumer compile — RIGHT, nothing breaks. git grep -w for the four names plus diagnoseViewMetadata and ViewMetadataDiagnosis. objectstack at head, outside packages/spec: nothing but api-surface / export-origins JSON; inside spec: index.ts, a build-schemas.ts comment, the three zod files and 8 test files. objectui at the pin f8a9d0fb and at origin/main 6fa5f64a: none of the four; only JoinedReportBlock (5 files), untouched and still unknown, so its inverted IsUnknown pin in report-chart-query-spec-parity.test.ts still holds (control at the pin: 15 files name ViewMetadataSchema / VIEW_METADATA_MEMBERS). cloud 48d7066 and origin/main 96eb092: none (control defineStack: 23 files). The Console Pin Gate is skipped at this head by its console path filter (.objectui-sha and console scripts only), so the census is the evidence, and it is clean.
• Declaration size / TS7056 — RIGHT. pnpm --filter @objectstack/spec build in the review worktree: exit 0, TS7056 ×0. Chunks: view.zod 500,541 B, action.zod 77,606 B, page.zod 305,845 B (the assembled aliases are emitted in the page.zod chunk); total .d.ts + .d.mts 30,836,475 B, byte-identical to the dev's readings. Aliases are emitted verbatim (InlineAction as z.input of the pipe's out; ViewMetadataParsed as z.infer of VIEW_METADATA_MEMBERS[ViewMetadataBranch]), no expansion.
• Pins — RIGHT, substantive, enforced. All three files are under src/** (in tsconfig.test.json's include) with no test-typecheck-debt.json entry, so any error reds check:test-typecheck. Narrowed tsc over the three pins plus the five test files that call diagnoseViewMetadata, under the test options: 8 errors, all in src/ui/view.test.ts, exactly the ledger's 4 signatures / 8 count; 0 elsewhere. In-memory ablation reverting each alias to its base spelling: InlineAction 4×TS2578, ViewMetadataParsed 3×, AssembledViewArtifact 3×, AssembledViewArtifactParsed 1×, so every @ts-expect-error fires. vitest on the three files: 3 files / 17 tests pass. Ledger diff against base: none; no pinned signature flipped.
• Corrected pending note — TRUE whichever release carries either entry. Head reads "ViewMetadataParsed is not changed by this change. It is re-derived from the same members, as their output types, by 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)." Both entries are pending on main once this merges and ship together; if a release consumes the fix(spec): ViewMetadata names a view body (the union of its members’ input types), not unknown #19919 entry first, main's deletion wins on merge and the sentence never ships. check-empty-changeset --base origin/main exit 1 names only this file (deliberate correction class), confirmed on the PR in 5857794016.

② Semver level

• minor + Clause-②: no (narrowing) + BREAKING banner + adr-0087: not-required (no-migration-prescription) — RIGHT, and identical in form to the merged fix(spec): ViewMetadata names a view body (the union of its members’ input types), not unknown #19919 changeset (119a02bc). check-changeset-no-major.mjs: launch-window convention, a breaking change ships as minor and the banner plus the disposition are the carriers; run at head, exit 0, "no major bump". check-adr-0087-registration.mjs --base origin/main: exit 0, reads [BREAKING+clause-②-narrowing] not-required (no-migration-prescription); the "If your code stops compiling" paragraph does not trip hasMigrationPrescription (prose, no framed rewrite), as fix(spec): ViewMetadata names a view body (the union of its members’ input types), not unknown #19919's did not. type-surface-only is closed by its predicate 2 (the diff touches packages/spec/**). fix: with minor raises the bump, never lowers it.
• FROM → TO per name — ACCURATE. Base e0f17a37 spellings were exactly z.input / z.infer of the schemas (the bare-alias population diff base → head removes exactly InlineAction and AssembledViewArtifact); head spellings match the table; "a container body is now a compile error here, as it always was at the schema" verified on both sides; "JoinedReportBlock … still resolves to unknown" verified (flags=2).
• Body versus changeset: the PR body's Clause-②: no lacks the (narrowing) arm the changeset carries. check-changeset-no-major's level axis reads the body: a bare no stands the axis down (exit 0); no (narrowing) would enforce and still pass, since the changeset is minor. Green either way, but fix(spec): ViewMetadata names a view body (the union of its members’ input types), not unknown #19919's body carried no (narrowing) and the seat's own ACCEPT names the arm, so a one-line body edit is owed by the seat before readying. Not blocking.

③ Boundary flags

• JoinedReportBlock: the remainder, untouched, still unknown at head (compiler flags); PR fix(spec): a joined report draws no chart — retire blocks[].chart and refuse a joined container chart (#20161) #20238 is still an open draft at 4aecf01d at this reading. The dev's measured fork (+15.8 KB report.zod.d.ts, no TS7056 with the annotation removed) is recorded on the PR for the taker; not re-measured here.
• ViewItem / ViewItemWire config: unknown — CONFIRMED by probe on the input and output of the viewItem member (viewItemArmShape(viewKind, config: z.ZodTypeAny); zod 4.6.1's ZodTypeAny defaults to unknown). So a viewKind: 'list' body with config: 42 type-checks as ViewMetadataParsed and AssembledViewArtifact while the door refuses it. Stays on the card as the second remainder, as the seat ruled.
• NEW, same family, named nowhere: both flattened-overlay members' static viewKind is 'form' | 'list' on input and output, because flattenedViewOverlayFields(kind: 'list' | 'form') is not generic and z.enum([kind]) widens to the parameter type. Effect: { type: 'grid', columns: ['name'], object, viewKind: 'form' } type-checks as AssembledViewArtifact, AssembledViewArtifactParsed, ViewMetadata and ViewMetadataParsed (through the list member) while both doors refuse it (list member: arm mismatch; form member: type and columns). Pre-existing since PR fix(spec)!: a flattened view overlay is judged by the member its viewKind names (#20186) #20245 (a91d12af), inherited by fix(spec): ViewMetadata names a view body (the union of its members’ input types), not unknown #19919's ViewMetadata too, and outside this claim's fence ("⛔ not the flattened-overlay members"). One-line fix in the fix(spec)!: a flattened view overlay is judged by the member its viewKind names (#20186) #20245 region: make flattenedViewOverlayFields generic in K extends 'list' | 'form'. Fold into the card's closing findings beside config: unknown. Not blocking.
• The .overwrite() list type default absent from the output type (①): sound but unnamed; a TSDoc sentence on ViewMetadataParsed / AssembledViewArtifactParsed closes it. Non-blocking.
• check:spec-parsed-alias: exit 0 at head; population 1443 → 1441 bare, 657 → 655 paired, because its STATE_ALIAS regex reads only the z.input of typeof X; spelling. Both …Parsed siblings exist; the gate no longer sees the pairs for InlineAction and AssembledViewArtifact, as for ViewMetadata since fix(spec): ViewMetadata names a view body (the union of its members’ input types), not unknown #19919. Acceptance note; a widened recogniser is a follow-up, not owed here.
• AssembledViewArtifactSchema accepts { name, object, viewKind: 'list' } (a content-less list patch) that ViewMetadataSchema refuses at the identity precondition, because the assembled union has no preprocess. Pre-existing runtime posture, untouched by this PR; observed, not filed.
• Advisory Check Changeset red: the deliberate correction of .changeset/view-metadata-type-not-unknown.md, confirmed in 5857794016. The job runs on pull_request only and is not among main's ruleset required contexts (TypeScript Type Check, Test Core, Dogfood Regression Gate, Build Core, Temporal Conformance (live PG + MySQL), Lint & Repo Gates, Governed Surface Queue Guard).
• CI at f8792c93, final state read 2026-09-27T17:04Z: 35 check-runs, 29 success, 5 skipped by filter (Console Pin Gate, Build Docs, Check PR Size, Auto Label, Packed-tarball smoke), 1 failure (Check Changeset, above), 0 running; every required context green (Test Core rollup completed success). main moved 7 commits past base e0f17a37, none touching this diff's files; git merge-tree against origin/main a9fb83ef is clean.

Implemented-by: claude/issue-19920-exported-types-unknown
Reviewed-by: session_01Rjy9MeetSfq34PKn81CRiN

VERDICT: PASS