docs(report-schema): drop the retired onSave / onCancel and put the Report Builder example under check:doc-snippets

[claude]

Fixes #7488

ReportBuilderSchema.onSave / .onCancel have been ?: never on the TypeScript face and
handlerKeyRefusal(...) arms on the zod mirror since PR objectui#7468 (objectui#7344, in the
objectui#6124 shape). The Report Builder example on content/docs/core/report-schema.mdx
still authored both, so the page taught a literal that is a tsc assignment error and a
named safeParse refusal. The two lines are removed — the disposition objectui#7340 chose
for seven of its eight pages.

The example is not hollow without them: the section teaches the ReportBuilderSchema shape
(type, report, availableFields, showPreview), not a save/cancel flow.

⭐ Why nothing went red, measured rather than asserted

Three instruments were green while the defect was present, each for its own reason:

instrument	why it could not see the line
check:doc-snippets	compiles ts / tsx fences only — this block was ```plaintext
check:doc-types	reads type string literals, nothing else
component-docs-retired-handler-keys-7340.test.ts	its DOC_MEMBER_RE requires key?: type (a member DECLARATION row) or a markdown prop-table row; the example's onSave: 'handleSaveReport', is an ASSIGNMENT and matches neither

⚠️ A fourth instrument did see the block, and that nuance matters for the class:
check:doc-fences classifies it as a SYNONYM finding (a TypeScript body under an
unhighlighted fence) and carries it inside a declared SHRINK-ONLY ledger — this page was
baselined at 8 such blocks, and --list names our block as one of them
(content/docs/core/report-schema.mdx:404). So the block being hidden from the compiler was
a known, baselined state, not an unknown one.

The reading the card asked for

Re-fenced the block ts with the two lines still present, then ran check:doc-snippets:

  [semantic]  content/docs/core/report-schema.mdx:421:3  TS2322: Type 'string' is not assignable to type 'undefined'.
  [semantic]  content/docs/core/report-schema.mdx:422:3  TS2322: Type 'string' is not assignable to type 'undefined'.
Semantic phase: 457 of 457 block(s) judged, 1 failed.

Exactly the two lines this card is about, and nothing else on the page — no other block on
this page uses an ambient name, none continues an earlier block. ⇒ per the dispatch ruling,
the re-fence lands too. After the deletion, on the same tree:

Covered blocks: 615 — 457 to compile, 158 declared fragment(s).
Semantic phase: 457 of 457 block(s) judged, 0 failed.
Every covered documentation snippet compiles against the built types.

456 → 457 compiled blocks: the page has joined the gate's scan surface, so this class is dead
on this page by construction — a retired key in a compiled block is a tsc error.

⚠️ A second file, and why it is not scope creep

scripts/check-doc-fence-languages.mjs, one line: content/docs/core/report-schema.mdx
8 → 7. The re-fence cannot land without it — that gate went red the moment the block left its
ledger, and it printed its own remedy:

❌  check:doc-fences — 1 stale KNOWN_UNHIGHLIGHTED_TS_FENCES entry/entries:
  content/docs/core/report-schema.mdx  — baselined at 8, now carries 7
    Good news, and the list has to say so: lower the numbers in
    KNOWN_UNHIGHLIGHTED_TS_FENCES in scripts/check-doc-fence-languages.mjs

That is the lane's documented protocol, stated in the gate's own header: "Every batch of that
lane now lowers these numbers in the same pull request that re-fences the blocks." The number
only ever goes down; no supported route raises it. Repo-wide the ledger moves 90 → 89 blocks.

⚠️ The pin that cites this page

packages/types/src/__tests__/handler-keys-string-any-mirrors-7344.test.ts cites this page's
onSave: 'handleSaveReport' as the corpus source of the retired dialect. Read before
deleting: the citation is a provenance comment, and the value is a hard-coded constant —

const AUTHORED_STRINGS = ['handleSaveReport', 'toast("Hello from ObjectUI!")'] as const;

— while every file that test actually reads comes from packages/types/src/zod (its
ZOD_DIR walk). It matches by existence-at-authoring-time, never by live content, so the
deletion cannot redden it, and it is green here. ⛔ The pin is not edited. The comment now
cites a page that no longer carries the spelling; that is a historical citation of where the
dialect came from and still reads correctly, so it is reported rather than rewritten.

Out of scope, filed rather than repaired

objectui#7840 — the objectui#7340 pin cannot see example object literals at all
(DOC_MEMBER_RE requires the ?). Widening it closes the class rather than this
instance, but it is its own measurement: the parse surface of example literals is far larger
than that of declaration rows, an owner has to be resolved from a type annotation for the
PAIR RULE to run at all, and the false-positive cost is unmeasured. ⛔ Not attempted here.

Evidence

Everything below ran on 9d46971b1, the head of this branch, with the doc gate's scoped
closure built first (pnpm exec turbo run build $(node scripts/check-doc-snippet-types.mjs --build-filter) --concurrency=2 — 34/34 tasks successful).

command	verdict
node scripts/check-doc-snippet-types.mjs	exit 0 — "Every covered documentation snippet compiles against the built types."
node scripts/check-doc-component-types.mjs	exit 0 — "✅ Every documented component type is registered."
node scripts/check-doc-fence-languages.mjs	exit 0 — "every TypeScript block in 227 document(s) is fenced ts/tsx/typescript, except 80 declared file(s) carrying 89 block(s)"
node scripts/check-doc-fence-languages.mjs --self-test	exit 0 — 26 cases pass
pnpm exec vitest run packages/types/ scripts/__tests__/check-doc-fence-languages.test.ts scripts/__tests__/check-skill-examples.test.ts scripts/__tests__/one-authority-per-exported-name-6273.test.ts	exit 0 — Test Files 130 passed (130), Tests 2384 passed (2384)
node scripts/check-doc-links.mjs	exit 0 — "Links are valid across 17 scan roots."
node scripts/check-docs-route-eager-closure.mjs	exit 0
node scripts/check-doc-example-shared-reader.mjs	exit 0
node scripts/check-handler-key-read-sites.mjs	exit 0
node scripts/check-control-bytes.mjs	exit 0 — 6365 tracked text file(s) scanned

The three test suites beyond packages/types are the ones that name
check-doc-fence-languages — a gate script's own suites are owed separately from the gates
its diff triggers.

node scripts/check-changeset-presence.mjs — exit 0, verdict quoted verbatim:

✅ No source or published contract of a released package changed in this range, so no changeset is owed.

⇒ no changeset file, and ⛔ no skip-changeset label: in this repository that label is read
by no workflow and exempts nothing; the gate's own verdict line is the authority.

Lint, narrowed and proven rather than skipped. Asked ESLint itself about the two changed
paths, through its own config resolution:

content/docs/core/report-schema.mdx     | ignored: true  | has rules: n/a
scripts/check-doc-fence-languages.mjs   | ignored: false | has rules: 0
packages/types/src/reports.ts (control) | ignored: false | has rules: 116

The flat config's only file selector is **/*.{ts,tsx}, so this diff intersects the linted
population in zero rules; the control shows the query measures something real. And
tseslint.configs.recommended is configured with no project / projectService, so linting
is not type-aware and this diff cannot move the verdict of any file it did not touch.

Not run, stated plainly: the full turbo run lint and the full test suite (CI runs both);
no browser verification (documentation-only change with no runtime surface); no reverse
ablation beyond the before/after pair above, which is itself the card's required measurement
— every instrument was captured green on the unmodified tree first, which is the point the
card makes.

Session, as prose because footers get rewritten: this branch was produced by the os-dev
seat at https://claude.ai/code/session_01KbJQ1y1J12nZxYzFWhP8Q3.

---

Generated by Claude Code