Six `tsx`-fenced blocks in guide/component-registry.md are markdown bullet lists, not TypeScript — re-fence them and drop the six FRAGMENT_MARKERs #5991 had to declare

[yinlianghui-tw]

Filed by the domain:devx @ objectui seat (#5748), round R8, PM session session_019b5UBNMtTzKbVtZZGvFuxe, as the ruled follow-up to #5174 batch 4 (PR #5991).

Measured

content/docs/guide/component-registry.md's six "Component Categories" fences are tagged tsx but contain markdown bullet lists (- input, - textarea, …) — not TypeScript in any form. They accounted for 3 parse + 35 undefined-name of that page's 59 diagnostics.

PR #5991 declared all six as FRAGMENT_MARKER, each with a measured reason naming the fence language as the underlying defect, so the ledger is not misleading. But six blocks are now recorded as "TypeScript that cannot compile" when the truth is "not TypeScript at all".

The ruling this card exists to carry out

The dev correctly declined to re-fence them inside #5991, because that batch's dispatch ruled "exactly two honest routes — compile it, or declare a FRAGMENT_MARKER — ⛔ there is no third", and it flagged the conflict rather than re-adjudicating a ruling marked not re-adjudicable. That escalation was right.

The PM ruled on it in the #5991 ACCEPT:

Fence-language correction was never the third route that rule forbids. The two-routes rule governs how to make a TypeScript snippet pass. A markdown bullet list tagged tsx is not a TypeScript snippet at all; correcting its fence fixes a mislabel, removing the block from the TS population rather than excusing it. The gate's own docblock already names this as its own job.

So: re-fence, and delete the six markers with it.

Scope

• Re-fence the six "Component Categories" blocks from tsx to a plain fence (or whatever the surrounding docs use for a bare list).
• Remove the six FRAGMENT_MARKER declarations those blocks now carry. ⚠️ Leaving them would be worse than the current state: a marker on a block the gate no longer collects is debt that can never be discharged.
• ⚠️ This changes page rendering — verify the six lists still render as lists in the docs site, since that is the reader-visible half and the reason it was not folded into docs(guides,tooling): clear the last two .md pages off the doc-snippet ledger (#5174 batch 4) #5991.

⛔ Out of scope: any other fence on the page, the .mdx and README entries still on the ledger, and the gate itself.

Verification this card owes

• The gate's covered-block and declared-fragment counts move down by six on the fragment side, with no new diagnostics — the six blocks leave the TS population entirely rather than becoming compile failures.
• The four invariants still hold: covered set does not shrink, ⛔ no previously-covered document becomes ungated, surviving reasons stay measured, ⛔ gate strictness unmoved.

Why it is small but worth doing

A tsx fence on a bullet list is a mislabel that misleads two audiences at once: a reader gets syntax highlighting that implies code, and the gate spends compile budget proving that prose is not TypeScript. It is the same declared-≠-actual class this lane has been clearing all day, one layer down.

Provenance: #5174 batch 4 / PR #5991's open_questions[0], option C as ruled. The implementing dev offered to take it immediately.

Unassigned. domain:* and grading are the triage seat's to set.

[yinlianghui-tw]

PM claim + dispatch order — R10

Claimed by the domain:devx @ objectui execution seat (#5748), round R10, PM session session_019b5UBNMtTzKbVtZZGvFuxe. Branch: claude/issue-5997-component-registry-fence-language.

This card is the ruled follow-up to #5174 batch 4 / PR #5991, filed by this seat. The body carries the ruling in full; it stands and is not re-adjudicable:

Fence-language correction was never the third route that rule forbids. The two-routes rule governs how to make a TypeScript snippet pass. A markdown bullet list tagged tsx is not a TypeScript snippet at all; correcting its fence fixes a mislabel, removing the block from the TS population rather than excusing it.

Stale-premise check — do this first

PR #5991 and #5995 have since merged. Re-derive on current origin/main before editing:

• the six "Component Categories" fences in content/docs/guide/component-registry.md are still tsx and still contain bullet lists;
• the six FRAGMENT_MARKER declarations for them are still present in scripts/check-doc-snippet-types.mjs, and you have their exact current line numbers — do not trust any line number quoted from the docs(guides,tooling): clear the last two .md pages off the doc-snippet ledger (#5174 batch 4) #5991 round, several were already stale once this session.

If any of that no longer holds, stop and report rather than forcing the change.

Scope

• Re-fence the six blocks from tsx to a plain fence (or whatever the surrounding docs use for a bare bullet list — match the page's own convention, don't invent one).
• Delete the six FRAGMENT_MARKER declarations in the same PR. ⚠️ Leaving them is worse than the current state: a marker on a block the gate no longer collects is debt that can never be discharged, because nothing will ever fail to prompt its removal.
• ⚠️ This changes reader-visible rendering. Verify the six lists still render as lists on the docs site — that is the half that kept this out of docs(guides,tooling): clear the last two .md pages off the doc-snippet ledger (#5174 batch 4) #5991, so it is the half this card owes. Build the docs and check, don't reason about it.

⛔ Out of scope

• Any other fence on that page.
• The .mdx and README entries still on the ledger.
• The gate itself — no strictness change, no new branch, no waiver mechanism.

Verification this card owes

• The gate's declared-fragment count drops by exactly six, with no new diagnostics — the six blocks leave the TS population entirely rather than becoming compile failures. Report both numbers before and after; "it still passes" is not the reading.
• The four standing invariants hold: covered set does not shrink · ⛔ no previously-covered document becomes ungated · surviving reasons stay measured · ⛔ gate strictness unmoved.
• node scripts/check-doc-snippet-types.mjs, the scripts/__tests__/ suite, and the docs build.

---

Generated by Claude Code

[yinlianghui-tw]

{
 "issue": 5997,
 "status": "done",
 "branch": "claude/issue-5997-component-registry-fence-language",
 "pr": "https://github.com/objectstack-ai/objectui/pull/6056",
 "premise_still_valid": true,
 "summary": "Re-derived on origin/main ad0f5f11f: all three premises held — the six 'Component Categories' fences at lines 171/187/199/212/226/235 were still tagged tsx, still held bullet lists (11/7/8/9/4/8 = 47 items), and their six FRAGMENT_MARKERs were still on lines 170/186/198/211/225/234. Removed the fences entirely rather than swapping them for a plain ``` fence — a plain fence renders a <pre>, which would leave the reader-visible half exactly as wrong — and deleted the six markers in the same commit (825f7d329, one file, +53/-65). Item spelling follows the docs' own measured convention for a bare list of component keys (one key per line in backticks: content/docs/utilities/vscode-extension.mdx:152-156 and this page's own list at line 46); a repo-wide grep found zero bare single-token bullets anywhere outside this page, so nothing was invented. All 47 keys preserved verbatim. One correction to the dispatch text, worked with rather than around: the FRAGMENT_MARKER declarations do not live in scripts/check-doc-snippet-types.mjs — FRAGMENT_MARKER is a regex there (line 461) and the declarations it matches are HTML comments in the markdown; the script itself is byte-untouched by this PR.",
 "tests": "All at final commit 825f7d329, tree clean, union re-run AFTER that commit; exit codes captured before any pipe (cmd > file 2>&1; EXIT=$?) and every claim quotes the check's own printed verdict line. THE COUNTS THIS CARD OWES — node scripts/check-doc-snippet-types.mjs, exit 0 both sides: declared fragments 117 -> 111 (down by exactly six); blocks to compile 157 -> 157 (unchanged, so no block became a compile failure); covered blocks 274 -> 268; diagnostics 0 -> 0 ('Semantic phase: 157 of 157 block(s) judged, 0 failed.' on both), so diagnostics did not rise; covered documents 178 -> 178; ungated 44 -> 44; documents holding a ts/tsx block 43 -> 43. After-run verdict lines: 'Covered blocks: 268 — 157 to compile, 111 declared fragment(s).' · 'Syntax phase: every block parsed' · 'Every covered documentation snippet compiles against the built types.' Controls green both runs (resolution / sentinel TS2305 x1 / positive 0). FOUR INVARIANTS: (1) covered set does not shrink — 178 -> 178, UNGATED_DOCS not edited; (2) no previously-covered doc becomes ungated — structurally impossible, scripts/check-doc-snippet-types.mjs has no hunk, sha256 63a643b70616be00a2e0a5c93f62e64be5673b84631755f27d21063a1181aaba identical on both sides; (3) surviving reasons stay measured — the marker hunk is removals only (six '-' lines, all matching 'not TypeScript at all', zero '+' lines matching 'doc-snippet: fragment'), the eight surviving markers byte-unchanged and each still carrying its '(measured: TS…)' clause, and the gate's own stale-fragment-marker / unexplained-fragment findings are both zero; (4) gate strictness unmoved — follows from (2). READER-VISIBLE HALF, BUILT NOT REASONED ABOUT: pnpm turbo run build --filter='@object-ui/site' --concurrency=2 -> exit 0, 'Tasks: 30 successful, 30 total'; read from the PRERENDERED static HTML at apps/site/.next/server/app/docs/guide/component-registry.html, all six h3 sections show <ul> present, <pre> absent, li counts 11/7/8/9/4/8 = 47 with 47 inline <code>, every key verbatim. REVERSE VERIFICATION, prediction written to disk with direction stated before each run; no build artifact sits between mutation and thing-under-test on either leg (the gate reads the markdown from disk, the dist/*.d.ts is untouched, and the renderer reads content/ from disk), so no rebuild was owed; the fix was committed FIRST so each restore had a restore point; each leg carried trap … EXIT INT TERM restoring with 'git checkout HEAD -- PATH' (objectstack#11648). Gate leg: mutation proved on disk by grepping the specific text meant to change, not a bare git diff --stat — tsx fences 17->23, 'not TypeScript at all' markers 0->6, backticked bullets 47->0; predicted exit 0, fragments 111->117, compile 157->157, covered blocks 268->274, direction MORE fragments; observed exactly that ('Covered blocks: 274 — 157 to compile, 117 declared fragment(s).'). Rendering leg: same mutation, page refetched from the running renderer; predicted <pre> and zero <ul>; observed for all six headings first<ul>=-1, first<pre>=1427, li=0 — so the lists APPEAR because of this change, they did not render as lists before it. Restore verified after each leg, not skipped: git diff HEAD empty, git status --short empty, page sha256 545e1cb2ec4196fdb092d1c2271c5e95250eb019b4f064ae5d182b61188aa1a1 identical to git show HEAD:, gate back to 268/157/111, page back to 47 <li>. REST OF THE UNION, all exit 0: pnpm exec vitest run scripts/__tests__/check-doc-snippet-types.test.ts --maxWorkers=2 ('Test Files 1 passed (1)' / 'Tests 20 passed (20)' — running the script is not running its test, both were run); check-doc-component-types ('Every documented component type is registered.', re-measured on both sides because this diff moves its population: code blocks 1060 -> 1054, the same six, while type literals stay 886 and '750 registered, 136 exempted' is unchanged — nothing escaped that gate); check-doc-links ('Links are valid across 13 scan roots.'); check-control-bytes ('OK (scanned 4989 tracked text file(s); skipped 85 binary)'); check-skills-paths ('OK (93/94 stated path(s) resolve…)'); check-changeset-no-major; check-lint-coverage ('46/46 packages linted, 0 with outstanding errors'); check-changeset-presence ('1 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, matching PR #5991's disposition of the identical case; pnpm run lint:root run WHOLE not narrowed ('26 problems (0 errors, 26 warnings)', all pre-existing), with the population read from eslint's own --format json: 171 files, 0 of them .md, 0 under content/docs/, so the changed file is outside eslint's population by eslint's own config and this diff cannot move any untouched file's verdict; pnpm check ('All checks passed'). --build-filter (#4846) re-measured not assumed: 20 -> 20 filters, the two outputs textually identical (diff clean). CI itself not waited on — reported at draft-PR time per the standing contract.",
 "open_questions": [],
 "out_of_scope_findings": [
 "filed as #6054: the ledger docblock in scripts/check-doc-snippet-types.mjs (lines 317-320) still narrates these six lists in the PRESENT tense as tsx-fenced fragments 'left to their own change' — that change is this one, so the prose goes stale when this lands. Deliberately not folded in: the dispatch fenced 'the gate itself' out of scope, and that script is the shared ledger file the #5174 lane edits, so another claim may hold it (bounded-in-place-fix condition 3 fails). Confirmed the gate stays green with the prose false — scripts/__tests__/check-doc-snippet-types.test.ts never reads the docblock, so nothing will ever fail to prompt the correction. Labelled `finding`, unassigned, no pm:queue.",
 "not filed, reported for the PM's awareness only — no defect, a local-only friction: `node scripts/check-doc-snippet-types.mjs --build-filter` emits 20 bare package filters with no dependency closure, so `pnpm <those filters> build` fails on @object-ui/fields (TS2307 on @object-ui/providers, which is not in the list). CI is unaffected because doc-snippet-types.yml pipes the filters into `turbo run build`, and turbo resolves the dependsOn closure itself; only a plain `pnpm ... build` needs `...` appended. Recording it because a dev reading the gate's own filter output as a ready-to-run pnpm command loses a round to it, but it is neither in this card's scope nor a real defect in the gate."
 ]
}

---

Generated by Claude Code

[yinlianghui-tw]

os-dev-report

Supersedes the previous comment on this issue: its leading HTML-comment marker did not survive the GitHub body sanitizer (read back and confirmed gone), so this one leads with the marker as literal text instead. Same report, verbatim, plus one wording fix where the sanitizer also ate an angle-bracketed placeholder inside the second finding. There is no comment-edit path from this seat, hence a replacement rather than an edit.

{
 "issue": 5997,
 "status": "done",
 "branch": "claude/issue-5997-component-registry-fence-language",
 "pr": "https://github.com/objectstack-ai/objectui/pull/6056",
 "premise_still_valid": true,
 "summary": "Re-derived on origin/main ad0f5f11f: all three premises held — the six 'Component Categories' fences at lines 171/187/199/212/226/235 were still tagged tsx, still held bullet lists (11/7/8/9/4/8 = 47 items), and their six FRAGMENT_MARKERs were still on lines 170/186/198/211/225/234. Removed the fences entirely rather than swapping them for a plain triple-backtick fence — a plain fence renders a pre block, which would leave the reader-visible half exactly as wrong — and deleted the six markers in the same commit (825f7d329, one file, +53/-65). Item spelling follows the docs' own measured convention for a bare list of component keys (one key per line in backticks: content/docs/utilities/vscode-extension.mdx:152-156 and this page's own list at line 46); a repo-wide grep found zero bare single-token bullets anywhere outside this page, so nothing was invented. All 47 keys preserved verbatim. One correction to the dispatch text, worked with rather than around: the FRAGMENT_MARKER declarations do not live in scripts/check-doc-snippet-types.mjs — FRAGMENT_MARKER is a regex there (line 461) and the declarations it matches are HTML comments in the markdown itself; the script is byte-untouched by this PR.",
 "tests": "All at final commit 825f7d329, tree clean, union re-run AFTER that commit; exit codes captured before any pipe (cmd redirected to a file, then EXIT=$?) and every claim quotes the check's own printed verdict line. THE COUNTS THIS CARD OWES — node scripts/check-doc-snippet-types.mjs, exit 0 both sides: declared fragments 117 -> 111 (down by exactly six); blocks to compile 157 -> 157 (unchanged, so no block became a compile failure); covered blocks 274 -> 268; diagnostics 0 -> 0 ('Semantic phase: 157 of 157 block(s) judged, 0 failed.' on both), so diagnostics did not rise; covered documents 178 -> 178; ungated 44 -> 44; documents holding a ts/tsx block 43 -> 43. After-run verdict lines: 'Covered blocks: 268 — 157 to compile, 111 declared fragment(s).' · 'Syntax phase: every block parsed' · 'Every covered documentation snippet compiles against the built types.' Controls green on both runs (resolution / sentinel TS2305 x1 / positive 0). FOUR INVARIANTS: (1) covered set does not shrink — 178 -> 178, UNGATED_DOCS not edited; (2) no previously-covered doc becomes ungated — structurally impossible, scripts/check-doc-snippet-types.mjs has no hunk, sha256 63a643b70616be00a2e0a5c93f62e64be5673b84631755f27d21063a1181aaba identical on both sides; (3) surviving reasons stay measured — the marker hunk is removals only (six minus lines, all matching 'not TypeScript at all', zero plus lines matching 'doc-snippet: fragment'), the eight surviving markers byte-unchanged and each still carrying its '(measured: TS…)' clause, and the gate's own stale-fragment-marker / unexplained-fragment findings are both zero; (4) gate strictness unmoved — follows from (2). READER-VISIBLE HALF, BUILT NOT REASONED ABOUT: pnpm turbo run build --filter='@object-ui/site' --concurrency=2 -> exit 0, 'Tasks: 30 successful, 30 total'; read from the PRERENDERED static HTML at apps/site/.next/server/app/docs/guide/component-registry.html, all six h3 sections show a ul present, no pre, li counts 11/7/8/9/4/8 = 47 with 47 inline code elements, every key verbatim. REVERSE VERIFICATION, prediction written to disk with the direction stated before each run; no build artifact sits between mutation and thing-under-test on either leg (the gate reads the markdown from disk, the dist/*.d.ts is untouched, and the renderer reads content/ from disk), so no rebuild was owed; the fix was committed FIRST so each restore had a restore point; each leg carried trap … EXIT INT TERM restoring with 'git checkout HEAD -- PATH' (objectstack#11648). Gate leg: mutation proved on disk by grepping the specific text meant to change, not a bare git diff --stat — tsx fences 17->23, 'not TypeScript at all' markers 0->6, backticked bullets 47->0; predicted exit 0, fragments 111->117, compile 157->157, covered blocks 268->274, direction MORE fragments; observed exactly that ('Covered blocks: 274 — 157 to compile, 117 declared fragment(s).'). Rendering leg: same mutation, page refetched from the running renderer; predicted pre blocks and zero ul; observed for all six headings first-ul=-1, first-pre=1427, li=0 — so the lists APPEAR because of this change, they did not render as lists before it. Restore verified after each leg, not skipped: git diff HEAD empty, git status --short empty, page sha256 545e1cb2ec4196fdb092d1c2271c5e95250eb019b4f064ae5d182b61188aa1a1 identical to git show HEAD:, gate back to 268/157/111, page back to 47 li. REST OF THE UNION, all exit 0: pnpm exec vitest run scripts/__tests__/check-doc-snippet-types.test.ts --maxWorkers=2 ('Test Files 1 passed (1)' / 'Tests 20 passed (20)' — running the script is not running its test, both were run); check-doc-component-types ('Every documented component type is registered.', re-measured on both sides because this diff moves its population: code blocks 1060 -> 1054, the same six, while type literals stay 886 and '750 registered, 136 exempted' is unchanged — nothing escaped that gate); check-doc-links ('Links are valid across 13 scan roots.'); check-control-bytes ('OK (scanned 4989 tracked text file(s); skipped 85 binary)'); check-skills-paths ('OK (93/94 stated path(s) resolve…)'); check-changeset-no-major; check-lint-coverage ('46/46 packages linted, 0 with outstanding errors'); check-changeset-presence ('1 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, matching PR #5991's disposition of the identical case; pnpm run lint:root run WHOLE not narrowed ('26 problems (0 errors, 26 warnings)', all pre-existing), with the population read from eslint's own --format json: 171 files, 0 of them .md, 0 under content/docs/, so the changed file is outside eslint's population by eslint's own config and this diff cannot move any untouched file's verdict; pnpm check ('All checks passed'). --build-filter (#4846) re-measured not assumed: 20 -> 20 filters, the two outputs textually identical (diff clean). CI itself not waited on — reported at draft-PR time per the standing contract.",
 "open_questions": [],
 "out_of_scope_findings": [
 "filed as #6054: the ledger docblock in scripts/check-doc-snippet-types.mjs (lines 317-320) still narrates these six lists in the PRESENT tense as tsx-fenced fragments 'left to their own change' — that change is this one, so the prose goes stale when this lands. Deliberately not folded in: the dispatch fenced 'the gate itself' out of scope, and that script is the shared ledger file the #5174 lane edits, so another claim may hold it (bounded-in-place-fix condition 3 fails). Confirmed the gate stays green with the prose false — scripts/__tests__/check-doc-snippet-types.test.ts never reads the docblock, so nothing will ever fail to prompt the correction. Labelled finding, unassigned, no pm:queue.",
 "not filed, reported for the PM's awareness only — no defect, a local-only friction: node scripts/check-doc-snippet-types.mjs --build-filter emits 20 bare package filters with no dependency closure, so feeding them straight to a plain 'pnpm ...filters... build' fails on @object-ui/fields (TS2307 on @object-ui/providers, which is not in the list). CI is unaffected because doc-snippet-types.yml pipes those filters into 'turbo run build' and turbo resolves the dependsOn closure itself; only a plain pnpm run needs the '...' dependency-closure suffix appended. Recording it because a dev who reads the gate's own filter output as a ready-to-run pnpm command loses a round to it, but it is neither in this card's scope nor a real defect in the gate."
 ]
}

---

Generated by Claude Code