docs(guides,tooling): clear the last two .md pages off the doc-snippet ledger (#5174 batch 4)

[yinlianghui-tw]

Part of #5174 — batch 4, the final batch. These were the last two content/docs.md entries on the ledger.

Ledger

	before	after
UNGATED_DOCS entries	46	44
covered documents	176	178 (strictly grows)
content/docs.md entries remaining	2	0
diagnostics cleared		104 = guide/component-registry 59 + guide/layout 45

The UNGATED_DOCS hunk contains only removals, so "zero previously-covered docs newly ungated" is structurally impossible to violate in this diff. Verified programmatically against d8b48f495's copy of the script: ADDED ledger entries: [] · previously-covered docs now UNGATED: [] · surviving entries whose REASON TEXT changed: [] · surviving entries with a non-measured reason: [].

What remains on the ledger is 12 .mdx pages and 32 package READMEs.

Routes — the two honest ones, no third

32 blocks brought under the gate: 9 compile (5 already did untouched, 4 after an edit) and 23 declared fragments with measured reasons.

page	blocks	compile	fragments
guide/component-registry.md	23	9	14
guide/layout.md	9	0	9

The real defect, fixed rather than declared

guide/component-registry.md's one complete copy-paste example — the RatingComponent the page invites a reader to take whole — calls useState while importing only forwardRef from react. It was broken at the first hook for anyone who copied it. The same block took cn from the reader's own @/lib/utils alias when @object-ui/componentsexports cn itself; sourcing it from the package makes the block compile and removes a dangling alias a reader may not have.

Three more blocks (getAllTypes / has / getMeta) were made self-contained by adding the one import a reader copying that block actually needs — the objectui#5047 disposition, applied to the document rather than to the harness.

What this green run does NOT mean

Stated because a green run must not be read as more than it is:

• ComponentRegistry is a Registry parameterised with any, so register()'s component argument is any. No registered component's props are checked against what SchemaRenderer passes.
• BaseSchema carries [key: string]: any, so a wrong key on a BaseSchema-shaped literal is structurally invisible to this gate. "No excess-property errors" is not "the keys are right".

What is sealed, and therefore really checked: ComponentMeta, ComponentInput, NavItem, NavGroup, AppShellProps, SidebarNavProps. Every literal of those six across both pages was measured clean — the four lucide-react blocks by probing them with the icon import shimmed, not by inferring it from the TS2307 that was hiding them. That probe is the batch-3 lesson applied: a broken import is a lid, so what sits under it gets measured, not assumed.

Declared, with the reason recorded rather than worked around

• lucide-react supplies the icon componentsNavItem.icon takes. It is a dependency of ten workspace packages (@object-ui/layout among them) but not a root dependency, so the specifier does not resolve in this gate's program. Four guide/layout.md blocks are declared for it. Hoisting a package to the repo root to buy a doc snippet coverage is a real dependency edge — the same call batch 3 made for class-variance-authority.
• guide/component-registry.md's six category lists are markdown bullet lists inside tsx fences. A leading - reads as unary minus on an undeclared name, and the form list does not even parse because switch is a keyword. The fence language is the underlying defect; correcting it changes the page's rendering rather than a snippet, which is neither of the two routes this batch is allowed, so it is declared here and recorded for its own change.

Gate strictness — proven byte-for-byte, not argued

Everything from the // ── Fence scanning banner to EOF (scanFences, listDocuments, derivePackageTypePaths, analyze, compileSnippets, reporting, main) is byte-identical to d8b48f495: 475 lines, sha256 dd9e52a11f116f3b on both sides. DOC_EXTENSIONS, TS_FENCE_LANGUAGES, FRAGMENT_MARKER, MIN_REASON_LENGTH and COMPILER_OPTIONS each hash identical too. The script diff is the two removed entries plus the ledger docblock's own prose.

--build-filter (#4846) — re-measured, not assumed

Filters 20 → 20, the two sets textually identical (diff clean). Turbo build tasks 33 → 33, identical package-by-package from --dry=json (tasks ADDED: [], tasks REMOVED: []). Covering these two pages costs this gate zero extra build tasks — the packages they import were already in the closure. Nothing for the maintainer to weigh.

Verification

All at final commit 1178f9a9a, working tree clean and byte-identical to what is pushed; the union was re-run after that commit. Exit codes captured before any pipe; every line quotes the gate's own printed verdict.

• node scripts/check-doc-snippet-types.mjs → exit 0: Scanned 222 document(s): 178 covered (42 of them hold a ts/tsx block), 44 ungated · Covered blocks: 273 — 156 to compile, 117 declared fragment(s). · Syntax phase: every block parsed · Semantic phase: 156 of 156 block(s) judged, 0 failed. · Every covered documentation snippet compiles against the built types. Controls green.
• pnpm exec vitest run scripts/__tests__/check-doc-snippet-types.test.ts → Test Files 1 passed (1) / Tests 20 passed (20). Running the script is not running its test; both were run.
• node scripts/check-doc-component-types.mjs → exit 0: Every documented component type is registered.
• node scripts/check-doc-links.mjs → exit 0 · node scripts/check-control-bytes.mjs → exit 0 · node scripts/check-skills-paths.mjs → exit 0 · node scripts/check-changeset-no-major.mjs → exit 0 · node scripts/check-lint-coverage.mjs → exit 0 (46/46 packages linted).
• node scripts/check-changeset-presence.mjs → exit 0: 3 file(s) changed, 0 of them under the src/ of a package the release covers … no changeset is owed.Verdict followed; none added, and no skip-changeset label (A skip-changeset label exists on this repo, but no workflow reads it — the repo's own test records that neither the workflow nor the label was ever real #4912/.github/WORKFLOWS.md documents 5 workflows that do not exist and omits 9 that do — including a changeset gate and a skip-changeset label neither of which is real #3724).
• pnpm run lint:root — the complete root eslint task, run whole, not narrowed → exit 0: 26 problems (0 errors, 26 warnings), all pre-existing; the changed .mjs reports 0 errors / 0 warnings. Population read from eslint's own --format json: 166 files, 0 of them .md, so both changed .md files are outside eslint's population by its own config; the other 46 lint tasks are per-package and this diff touches nothing under packages/, apps/ or examples/.
• pnpm check → exit 0: All checks passed.

Reverse verification

Prediction written to disk before each run with the direction stated. Build artifact between mutation and thing-under-test: none on either leg — the mutation is markdown the gate reads straight from disk and the dist/*.d.ts it compiles against is untouched — so no rebuild was needed. Each leg proved the mutation reached disk by grepping the specific text meant to change, and carried trap … EXIT INT TERM restoring with git checkout HEAD -- PATH (objectstack#11648: git checkout REF -- PATH writes the index, so a trap restoring with a bare git checkout -- PATH reads the mutation back and exits 0).

• Leg 1, revert component-registry.md only, ledger removal kept. Predicted exit 1 · exactly 59 diagnostics all on that page · fragments 117 → 103 · blocks-to-compile 156 → 170 · 1 syntax failure · NOTE printed · direction MORE. Observed exactly that, including the three named lines (TS2304 'BaseSchema', TS2307 './HeavyComponent', TS2304 'useState').
• Leg 2, revert both pages. Predicted exit 1 · exactly 104 split 59 / 45 per page · fragments 117 → 94 · blocks-to-compile 156 → 179 · covered blocks 273 unchanged · 3 syntax failures · direction MORE. Observed 104 split exactly 59 component-registry / 45 layout — matching the baseline to the unit and per page.
• Restore verified after each leg, not skipped: git diff HEAD empty and git status --short empty, 23 markers back on disk, gate exit 0 with 156 of 156 block(s) judged, 0 failed.

---

Generated by Claude Code