No gate reads inside a code fence in skills/objectui/ — 123 TypeScript fences and 63 JSON fences are checked by nothing

Filed out-of-scope from the #7251 skills-optimization flight (its DEFER row OUI-E-01). Unassigned.

Measured

At the flight's audit ref, skills/objectui/ carried 123 TypeScript fences and 63 JSON/JSONC fences. grep -rc "os:check" skills/objectui/ returns zero.

Every objectui gate that could read them excludes the surface, by construction rather than by accident:

Gate	Why it does not reach a fence in `skills/`
`check-skills-paths.mjs`	Deliberately excludes fences — "A fence is a worked example", and a tutorial fence may name a file the reader is about to create. It reads inline code spans in prose only.
`check-doc-links.mjs`	`SCAN_ROOTS` has no `skills` row.
`check-doc-component-types.mjs`	`DOCS_ROOT = 'content/docs'`; its header says so verbatim — "This gate walks `content/docs` and nothing else: not `skills/**`".
`check-doc-snippet-types.mjs`	`content/docs` + package READMEs.
`check-doc-fence-languages.mjs`	`content/docs` + package READMEs. It judges the fence language tag, never the contents.

Net: this repo enforces path existence in skill prose, and nothing else. A fence can import a symbol that does not exist, spell a key the renderer never reads, or restate a shape the spec changed, and stay green forever.

What already exists to copy

objectstack has packages/spec/scripts/check-skill-examples.ts, with the convention that an HTML comment containing os:check, placed on the line above a fence, opts that fence into a real type-check. A port would give objectui the same opt-in marker plus a runner over skills/**.

Why this is the highest-leverage structural addition here

It is the mechanism that stops the drift class from recurring. Two rounds of it have already been cleaned by hand — #3658 / #7094's truth sweep, and #7251's own correctness items, which included an eval asserting a mobile hook API (useSwipe / useLongPress) that has zero occurrences repo-wide. Every one of those was found by a human reading, which is precisely the economics check-skills-paths.mjs was created to end for paths.

Shape of the work

Port the marker convention and the runner; opt-in, so it starts at zero red.
Measure how many of the 123 fences compile as-is, and land the marker on the ones that already pass — that number is the honest starting ratchet.
Decide separately whether the ratchet should be shrink-only, the way check-doc-fence-languages.mjs treats its declared-file population.

Related: #7358 (the same package's other unguarded surface — the paths gate's scan root).

Design input from implementing #7360 (PR #7408), not a request to change scope here.

The oracle this check picks is load-bearing, and the two candidates disagree on a real row today.

#7360's audit measured each must_contain token against guides/BASENAME.md — the guide matching the eval file's own basename. That per-guide oracle is what found its rows. But the "teach the thing" half of that card is closed by adding one bullet to guides/plugin-development.md naming PluginScope.registerComponent, and three assertions require that token: evals/plugin-development.json evals 1 and 2, and evals/page-builder.json eval 1.

After PR #7408 the counts are:

registerComponent in guides/plugin-development.md: 0 to 1
registerComponent in guides/page-builder.md: 0, unchanged

So evals/page-builder.json eval 1 is honest under a bundle-wide oracle (the answering agent loads the whole skill, router and all guides) and still untaught under a per-guide one. Whichever this check adopts, that row is either fine or red — worth deciding deliberately rather than discovering when the gate first runs.

Second input: whole-token comparison changes what the check can see. #7360's body already raises this; two measured examples, both in guides/testing.md, both counting 1 rather than 0 so a substring check reports them clean:

create — matches create: vi.fn().mockResolvedValue({ id: '1', name: 'New' }), a mocked dataSource method, where the eval means the form mode
view — matches the comment // vite preview

A third, FieldWidgetProps, matched only the head of FieldWidgetPropsSchema and is corrected in PR #7408. These three are the argument for comparing whole tokens; they are filed as rows in #7405 so they are not lost if this check lands with substring semantics.

Third, smaller input: must_not_contain deserves the opposite treatment. #7370 deliberately moved a token there as "props": { rather than the bare word, because a bare props fails any answer that names the rule while following it. A check that validates must_not_contain entries against the guides the way it validates must_contain would get that backwards.

Generated by Claude Code

Claim: skills lane seat — graded and dispatched in one write (priority:p2 + pm:dispatched; domain:skills already held). Triage is vacant in objectui and this is the lane's own programme follow-up (filed from #7251's DEFER row OUI-E-01), so the seat grades it. Value gate (#13597): it is the mechanism that ends a drift class two hand-sweeps have already paid for — passes.
Session: session_01LraLgQVGq8egUwfYZpbYt1
Branch: claude/issue-7359-skill-fence-check
Worktree: objectui-issue-7359
Clause-②: no — a gate, no contract claim. Governed: the marker landing touches skills/objectui/** ⇒ the PR stays DRAFT (ui#6866 批 #7); review requests are the seat's step.

Scope, exactly the card's three steps with step 3 recorded not decided:

Port the  opt-in convention from objectstack's packages/spec/scripts/check-skill-examples.ts (marker on the line directly above a fence) as an objectui runner over skills/objectui/**/*.md — TypeScript/TSX fences type-checked, JSON/JSONC fences parsed — with a --self-test, a scripts/__tests__/ spec, a package.json script, and a workflow modelled on skills-paths.yml (no paths / paths-ignore, for the reason that file states; its test pins that).
Measure how many fences compile as-is at the branch point and land the marker on exactly those — the honest starting count, stated in the PR body. Opt-in ⇒ zero red on day one.
⛔ Whether the marked population becomes shrink-only (the way check-doc-fence-languages.mjs treats its declared files) is NOT decided in this flight — record the number, add no ratchet.

Re-measured at objectui origin/main bf244f4, 2026-09-03T01:22Z: 112 ts/tsx/typescript fences and 56 json/jsonc fences across skills/objectui/**/*.md (the card's 123 / 63 were at its audit ref); os:check occurs 0 times. ⛔ The eval-token oracle discussed in comment 5514262423 (must_contain tokens vs the guides, whole-token vs substring) is a different check and is not this card — the seat records it separately. No guide prose is edited by this flight beyond the marker lines.

Generated by Claude Code

Claim: dev seat (os-dev subagent of the dispatching PM session — same session id, distinct seat)
Session: session_01LraLgQVGq8egUwfYZpbYt1
Branch: claude/issue-7359-skill-fence-check
Worktree: /home/user/objectui-issue-7359
Branch point: 75e7436 (objectui origin/main at worktree creation — main moved on from the bf244f4 the seat re-measured at, so the starting population is re-measured at this sha).

Scope taken is exactly the seat's three steps in comment 5518884588: port the opt-in marker convention + a runner over skills/objectui/**/*.md, wire CI, and land the marker on exactly the fences that pass as-is — number recorded, no ratchet. The eval-token oracle in comment 5514262423 is a different check and is not built here.

Generated by Claude Code

os-dev-report

{
  "issue": 7359,
  "repo": "objectui",
  "status": "done",
  "branch": "claude/issue-7359-skill-fence-check",
  "branch_point": "75e7436",
  "head_sha": "70fdb0e3fbc3e876347480c4db959a7743934523",
  "pr": "https://github.com/objectstack-ai/objectui/pull/7427",
  "pr_state": "draft — stays draft (governed skills/objectui/**, ui#6866 批 #7)",
  "population": {
    "ts": { "passing": 17, "total": 112 },
    "json": { "passing": 39, "total": 56 }
  },
  "markers_landed": 56,
  "import_resolution": "Built dist/*.d.ts, not tsconfig.base.json path mappings to src. A marked ts fence compiles --strict through the paths map check-doc-snippet-types.mjs derives from every package's own exports/types, plus that gate's declared-dependency rule for third-party specifiers; the type-check harness itself (derivePackageTypePaths, deriveDeclaredDependencyPaths, compileSnippets) is IMPORTED from that gate rather than re-rolled, so the syntax/semantics split and its four self-controls are shared. src was rejected for the reason that gate's header states: the root tsconfig maps the workspace to source, so an inherited config would judge the guides against code no reader resolves. Declared precondition: exit 2 PRECONDITION NOT MET on an unbuilt or source-typed tree, on a failed harness control, and on an EMPTY marked population — never exit 0, never exit 1. Honest edge, printed on every run as an 'Unmapped specifiers:' line: a bare specifier the repository ROOT declares still resolves through pnpm's root node_modules symlinks (in the marked population: vitest, @playwright/test); the inherited UNDECLARED control bounds transitives, which is a different leak. Recorded as follow-up 3 in the PR body, not silently absorbed.",
  "ci_wiring": "Own workflow .github/workflows/skill-examples.yml, no paths / paths-ignore, modelled on skills-paths.yml's trigger shape and doc-snippet-types.yml's install/derive-filter/build body. lint.yml was NOT the route: 'pnpm check' in this repo is 'node packages/cli/dist/cli.js check' (the product's own command), so there is no check:* aggregation to join. A second step inside skills-paths.yml was also rejected — that file's own header asks it stay a checkout plus one node call, and this gate needs a build. Classified rather than left to default into silence: 'Skill Example Check' added to REQUIRED_CONTEXTS in scripts/dependabot-merge-gate.mjs, merge_group subscribed, MUST_SUBSCRIBE_MERGE_GROUP reason added, and content/docs/guide/ci-cd-pipeline.md gained an inventory row plus a section (ci-cd-pipeline-doc.test.ts requires a heading naming every workflow file).",
  "premise_still_valid": true,
  "premise_false": [],
  "reverse_verification": [
    "LEG A marked ts fence: useObjectTranslation -> useObjectTranslationNoSuchExport in guides/i18n.md; mutation proved on disk by counting BOTH texts (removed 7 -> 0, injected -> 7); gate exit 1 with '[semantic] skills/objectui/guides/i18n.md:40:10 TS2305: Module \"@object-ui/i18n\" has no exported member' and 'Semantic phase: 17 of 17 ts fence(s) judged, 1 failed'; restored by 'git checkout HEAD -- PATH', proved by blob hash fc0d4ad31577ebc079bab867f88f49ac07f02f36 plus empty git diff HEAD.",
    "LEG B marked json fence: trailing comma injected in rules/styling.md (removed 1 -> 0, injected -> 1); gate exit 1 with '[json] skills/objectui/rules/styling.md:183 Expected double-quoted property name in JSON at position 86' and 'JSON phase: 39 fence(s) parsed, 1 failed'; restored, blob fe511a6f52568fe0f017c05f1ea4c216101ad904, git diff HEAD empty.",
    "LEG C marker adjacency: one blank line inserted between the marker and its fence in rules/composition.md (marker-followed-by-blank occurrences -> 1); gate exit 1 with 'skills/objectui/rules/composition.md:203 [orphan-marker]'; restored, blob 775ca26e706e43a32cfc7cb6bc27c126b9ea2c39, git diff HEAD empty.",
    "Restored-tree re-measure: gate exit 0, worktree clean, all three blob hashes match HEAD. trap on EXIT INT TERM, absolute paths throughout. NO build step sits between mutation and measurement — the artifact the gate reads IS the mutated markdown, so no dist can hold a stale copy; the dist-preflight rule does not apply and is not claimed.",
    "OPT-IN half measured, not assumed: on the same tree --measure reports 12 parse + 83 semantic + 17 json failures across all 168 candidates while the gate itself reports 0 failed and exits 0."
  ],
  "gates": [
    { "cmd": "node scripts/check-skill-examples.mjs", "exit": 0, "verdict": "Every marked skill example holds up against the built types. (Semantic phase: 17 of 17 ts fence(s) judged, 0 failed; JSON phase: 39 fence(s) parsed, 0 failed)" },
    { "cmd": "node scripts/check-skill-examples.mjs --self-test", "exit": 0, "verdict": "check-skill-examples self-test: 19 cases pass (marker adjacency both directions, orphans, nested illustration, json/jsonc, and the compiler legs through the real harness)" },
    { "cmd": "node scripts/check-skill-examples.mjs --measure", "exit": 0, "verdict": "Starting population — ts: 17/112 pass; json: 39/56 pass." },
    { "cmd": "pnpm exec vitest run (12 wiring/pin suites, --maxWorkers=2, under the shared verify lock)", "exit": 0, "verdict": "Test Files 12 passed (12); Tests 318 passed (318) — check-skill-examples, check-skills-paths, check-doc-snippet-types, merge-queue-reporting, dependabot-merge-gate, ci-cd-pipeline-doc, workflow-cache-save-bound, lint-workflow, scripts-type-check, entry-guard-wiring, check-doc-fence-languages, check-shell-escape-residue" },
    { "cmd": "node scripts/check-skills-paths.mjs", "exit": 0, "verdict": "check-skills-paths: OK (88/89 stated path(s) resolve across 20 guide file(s); 1 baselined)" },
    { "cmd": "node scripts/check-doc-links.mjs", "exit": 0, "verdict": "ran clean" },
    { "cmd": "node scripts/check-control-bytes.mjs", "exit": 0, "verdict": "ran clean" },
    { "cmd": "node scripts/check-shell-escape-residue.mjs", "exit": 0, "verdict": "check-shell-escape-residue: OK (5/5 root(s) resolved -- skills: 16 file(s), 196 fence(s); 207 file(s) and 1304 fenced block(s) examined in total) — the marker lines do not disturb its fence walk" },
    { "cmd": "node scripts/check-doc-fence-languages.mjs", "exit": 0, "verdict": "check:doc-fences — every TypeScript block in 225 document(s) is fenced ts/tsx/typescript, except 80 declared file(s) carrying 90 block(s) (SHRINK-ONLY)" },
    { "cmd": "node scripts/check-doc-snippet-types.mjs", "exit": 0, "verdict": "Every covered documentation snippet compiles against the built types. (Semantic phase: 417 of 417 block(s) judged, 0 failed)" },
    { "cmd": "node scripts/check-changeset-presence.mjs", "exit": 0, "verdict": "No source or published contract of a released package changed in this range, so no changeset is owed. — its own verdict decides; NO changeset added" },
    { "cmd": "node scripts/check-lint-coverage.mjs", "exit": 0, "verdict": "ran clean" },
    { "cmd": "node scripts/check-governed-queue-guard.mjs --self-test", "exit": 0, "verdict": "OK check-governed-queue-guard self-test: 132 cases pass" },
    { "cmd": "node scripts/check-governed-queue-guard.mjs --test CHANGE-SET-PATHS", "exit": 3, "verdict": "GOVERNED — the correct verdict for this change set, not a red: 'Do not flip it ready, enqueue it, or arm auto-merge. Park it as a DRAFT'" },
    { "cmd": "pnpm exec eslint --format json (2 changed lintable files)", "exit": 0, "verdict": "2 files linted, 0 errors, 0 warnings. Narrowing declared: pnpm lint is turbo run lint = each workspace package's own 'eslint .', and scripts/ has no package.json so it is in no package's lint run at all (the same structural gap scripts-type-check.test.ts records for type-check). Invariance: eslint.config.js configures NO type-aware linting (no project, no projectService anywhere in it), so this diff cannot move the verdict for any untouched file." }
  ],
  "skills_surface_budget": {
    "changed_files": 13,
    "per_file_lines_before_after": "SKILL.md 136/137; auth-permissions 345/348; data-integration 443/451; i18n 185/187; mobile 166/168; page-builder 321/331; plugin-development 442/446; project-setup 340/345; schema-expressions 511/518; testing 341/345; composition 277/278; protocol 314/320; styling 316/319",
    "whole_bundle_lines_before_after": "4356 -> 4412 (+56)",
    "note": "Every added line is one byte-identical inert marker comment. 56 insertions, 0 deletions, zero prose added or rewritten, zero fence content edited."
  },
  "follow_ups_recorded_not_taken": [
    "Shrink-only ratchet on the marked population — card step 3, explicitly deferred; number re-derivable with --measure.",
    "Bare `any` inside a marked block (objectstack's third assertion) — a new assertion rather than the convention, so not in this flight.",
    "Bound root-devDependency resolution the way the UNDECLARED control bounds transitives.",
    "Widen SCAN_ROOTS to .claude/skills/** the way check-skills-paths.mjs was widened by ui#7358."
  ],
  "out_of_scope_findings": [
    "REPORT ONLY, no card filed per dispatch. Two pin suites FAIL on a BUILT tree, and CI never sees it because Test (shard N/4) runs unbuilt: scripts/__tests__/check-sdui-registration-pins.test.ts derives 'packages/app-shell/dist/console/connect/ConnectAgentWidget.js' where it asserts the 'packages/app-shell/src/console/connect/ConnectAgentWidget.tsx' path, and scripts/__tests__/check-readme-exports.test.ts gets census.excerptsNotJudged 1 where it asserts 3. Both suites' own describe blocks claim to hold 'in BOTH build states', so this is the claim being false rather than a missing case. Unrelated to this change set, which touches zero packages/** files.",
    "REPORT ONLY. 17 of the 56 json fences in skills/objectui/ are tagged `json` while carrying `//` comments or `...` elisions (12 in guides/schema-expressions.md, 4 in rules/protocol.md, 1 in rules/styling.md). They are legitimately unmarked here, but `jsonc` would be the honest tag; check-doc-fence-languages.mjs judges fence language tags and does not scan skills/, so nothing says so today.",
    "REPORT ONLY. The root-devDependency resolution edge described in import_resolution is a property of the SHARED harness, so check-doc-snippet-types.mjs carries it too: its UNDECLARED control proves resolution did not widen to transitives, and says nothing about specifiers the repository root itself declares."
  ],
  "mcp_calls": 7
}

Note on the card's measured numbers: the card body states 123 TS / 63 JSON at its audit ref; the seat re-measured 112 / 56 at bf244f4 in comment 5518884588. This flight branched from 75e7436 (main had moved on again) and measured exactly 112 / 56 there, so the seat's re-measure holds at the branch point too. os:check occurred 0 times before this PR. The card's premise is intact — nothing was falsified.

Generated by Claude Code

ACCEPT — skills-lane seat (session session_01LraLgQVGq8egUwfYZpbYt1), reviewer of record, verified against GitHub and the seat's own objectui checkout, not the report.

PR feat(gates): check the marked fenced examples in the skill guides #7427 exists, is DRAFT, head 70fdb0e, 20 files, +1,793 −0, one commit; auto-labels configuration · ci/cd · dependencies · tests; no changeset — check-changeset-presence.mjs's own verdict ("No source or published contract of a released package changed in this range") decides, as briefed. mergeable_state: behind (branch point 75e7436; main has moved) — a draft awaiting a human merge, which brings main in.
Governed surface, measured by the seat at 2026-09-03T02:00Z: the only edits under skills/objectui/** are 56 byte-identical marker lines — 0 non-marker additions, 0 deletions across the 13 guide/rule files; marker count 56 at head, 0 on the base; every marker sits directly above a fence (the seat's own adjacency walk printed nothing). No prose, no fence content, no eval touched.
Design accepted as the card's step 1: the opt-in os:check convention ported from objectstack's runner with strict adjacency (an orphan marker fails), fence-awareness, and the type-check harness IMPORTED from check-doc-snippet-types.mjs (built dist/*.d.ts, the consumer's surface) rather than re-rolled; exit 0 / 1 / 2 with PRECONDITION NOT MET on an unbuilt tree, a failed harness control, or an EMPTY marked population (seat spot-checked scripts/check-skill-examples.mjs:145, :248, :516, :754) — a gate that checks nothing never reports success. The honest edge (root-declared bare specifiers such as vitest still resolve) is printed on every run, not hidden.
Step 2, the starting population: --measure at head — ts 17 / 112, json 39 / 56 pass; exactly those 56 are marked. Step 3 (shrink-only) correctly NOT decided; the number is re-derivable.
CI wiring: a dedicated skill-examples.yml on pull_request / push / merge_group with no paths / paths-ignore (seat-verified), pinned by its test the way skills-paths.yml is; Skill Example Check added to REQUIRED_CONTEXTS in scripts/dependabot-merge-gate.mjs, merge-queue-reporting.test.ts and the ci-cd-pipeline.md inventory updated. The lint.yml route was correctly rejected (pnpm check there is the product's own command).
Reverse verification, three legs with on-disk proof: a marked ts fence broken → TS2305 naming file:line; a marked json fence broken → a parse error naming file:line; a blank line between marker and fence → [orphan-marker]; each restored by blob hash + empty git diff HEAD; the opt-in half measured (--measure finds 112 failures across all 168 candidates while the gate exits 0).
CI at 70fdb0e: every one of the 28 check runs is success or skipped, read at 2026-09-03T02:12Z — including the new gate's own first run, Skill Example Check success (completed 02:01:01Z), Lint, Type Check, all four Test shards, Doc Snippet Type Check, Skill Guide Path Check, Shell Escape Residue Scan, Governed Surface Queue Guard (draft — correct), Changeset Declaration.
Follow-ups recorded on the seat ledger, not filed now: the shrink-only decision (step 3); the bare-any assertion port; bounding root-devDependency resolution; widening to .claude/skills/**; 17 json fences that are really jsonc (comments / elisions) — a tag question no gate reaches in skills/; two pre-existing pin suites failing on a BUILT tree (check-sdui-registration-pins, check-readme-exports) that CI's unbuilt shards never see — the seat dedups before filing. The eval-token oracle (comment 5514262423) remains its own future gate, as the claim said.
Governed skills/objectui/** ⇒ stays DRAFT; review requested from os-zhuang + hotlong; human merge is the record. At MERGED the seat strips pm:dispatched (Fixes ⇒ the card closes).

Generated by Claude Code

Landed — skills-lane seat (session session_01LraLgQVGq8egUwfYZpbYt1), merge patrol 2026-09-03T06:55Z (git log --grep '(#7427)' on a freshly fetched origin/main): PR #7427 merged as c92dc41 at 06:27:06Z — scripts/check-skill-examples.mjs, the Skill Example Check workflow and its REQUIRED_CONTEXTS row are on main; the card closed by the merge at 06:42:22Z (os-zhuang). pm:dispatched stripped; priority:p2 · domain:skills kept, read back. The follow-ups the flight recorded (the shrink-only decision on the marked population, the eval-token oracle as its own gate, the jsonc fences, the two pre-existing pin suites that fail only on a built tree, the bare-any assertion port, the root-devDependency resolution bound, widening to .claude/skills/**) are filed by the seat as separate cards after dedup, this check-in.

Generated by Claude Code

Navigation Menu

Repository navigation

No gate reads inside a code fence in `skills/objectui/` — 123 TypeScript fences and 63 JSON fences are checked by nothing #7359

Measured

What already exists to copy

Why this is the highest-leverage structural addition here

Shape of the work

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

No gate reads inside a code fence in skills/objectui/ — 123 TypeScript fences and 63 JSON fences are checked by nothing #7359

Description

Measured

What already exists to copy

Why this is the highest-leverage structural addition here

Shape of the work

Activity

os-litant commented on Sep 2, 2026

os-litant commented on Sep 3, 2026

os-litant commented on Sep 3, 2026

os-litant commented on Sep 3, 2026

os-litant commented on Sep 3, 2026

os-litant commented on Sep 3, 2026

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

No gate reads inside a code fence in `skills/objectui/` — 123 TypeScript fences and 63 JSON fences are checked by nothing #7359