test(scripts): give 30 more self-tests a battery roster and floor

[claude]

Part of #13799 (batch 1c, Tier B — batch-1 shape)

PR #13487's roster-and-floor shape, exactly as batch 1 landed it in PR #14851, transplanted onto the scripts/** self-tests that already carry that shape. 432 batteries / 2591 cases pinned across 30 files.

Per file, and nothing else:

• a module-level frozen SELF_TEST_BATTERIES roster of names with a per-battery floor, SELF_TEST_BATTERY_FLOOR pinning the roster's own size, and UNATTRIBUTED_BATTERY, placed after the import block (batch 1b's uniform placement — TDZ-safe by construction);
• battery('BANNER TEXT') at each existing section banner, named verbatim from the banner;
• one registerCase() at the head of the existing in-body assertion helper;
• the floor evaluated immediately before the existing verdict site, its messages entering the file's own failure sink so the report the file already prints carries them.

⛔ No case is rewritten, none reordered, no assertion changes meaning. ⛔ No TOTAL is pinned.

The file surface: 30 floored, 15 dropped with the measured reason

The dispatch named 45 files (the batch-1b report's "HAS a block-bodied in-body assertion helper" list). That list was derived on the helper half of the batch-1 criterion only. Re-measured against the tree on all three halves — a block-bodied in-body assertion helper and at least two named section banners and a body-level verdict site — 30 qualify.

The banner criterion, as measured

A named section banner is a comment block at statement position inside the self-test body (innermost open bracket is a brace, never a square bracket or a paren) whose head carries a section marker: a horizontal rule, a box (a bare rule line, then the title, then a rule), or an enumerator (// 1. name, // 3 — name, // G1: name, // TSO-6048: name). A comment with no marker is a paragraph explaining the next statement, not a section head; reading one as a section head is a judgement per comment, which a verbatim transplant does not make.

The bracket half of that rule is load-bearing and was measured: in scripts/docs-audit/affected-docs.mjs a brace-only depth count put four banners inside an array literal, where a battery() statement is a syntax error — the Tier-C shape the batch-1b report names. The probe pass caught it; the renderer now refuses that position.

Floored (30)

file	batteries	cases
scripts/check-adr-0087-registration.mjs	48	292
scripts/check-auth-mount-ledger.mjs	18	42
scripts/check-ci-filter-parity.mjs	7	42
scripts/check-comment-mask-adoption.mjs	6	23
scripts/check-console-injection.mjs	10	23
scripts/check-cross-repo-closer-outcome.mjs	5	81
scripts/check-dev-prereqs.mjs	16	51
scripts/check-dispatcher-error-vocabulary.mjs	2 (first battery hoisted)	282
scripts/check-doc-frontmatter.mjs	11	85
scripts/check-docs-locale-catch-all.mjs	15	28
scripts/check-docs-section-name.mjs	22	85
scripts/check-engine-split-ratio.mjs	2	16
scripts/check-error-status-conformance.mjs	23	40
scripts/check-i18n-stale-fill.mjs	2 (first battery hoisted)	27
scripts/check-init-service-contract.mjs	19	26
scripts/check-keyed-text-bounds.mjs	9	67
scripts/check-merge-queue-triage-outcome.mjs	7	126
scripts/check-overlay-whitelist-table.mjs	19	70
scripts/check-quick-reference-counts.mjs	22	37
scripts/check-ratchet-remedy-authority.mjs	20	27
scripts/check-required-contexts.mjs	22	150
scripts/check-route-envelope.mjs	27 (first battery hoisted)	88
scripts/check-section-landing-index.mjs	12	31
scripts/check-stall-guard-budget.mjs	10	61
scripts/check-system-context-census.mjs	17	43
scripts/check-undeclared-dep-imports.mjs	9	50
scripts/check-vendor-version-stamps.mjs	6 (first battery hoisted)	64
scripts/docs-audit/affected-docs.mjs	28 (first battery hoisted)	545
scripts/release-github-releases.mjs	10	49
scripts/sync-template-versions.mjs	8	40

Five files run cases before their first banner (check-dispatcher-error-vocabulary, check-i18n-stale-fill, check-route-envelope, check-vendor-version-stamps, docs-audit/affected-docs); their first battery is opened at the top of the body and that banner then carries no second opener — the PR #13487 precedent's own shape, as batch 1b did it. Banners that registered no case are not declared as batteries and get no opener: a battery is a section that actually runs cases.

Dropped (15) — left for a later batch rather than improvised

file	measured reason
scripts/check-declaration-mirrors.mjs	the in-body assertion sink is a CONCISE arrow — const ok = (label, cond) => cases.push({ label, cond }); (L411) — with no block body to hold registerCase(). That is the class-1 blocker, not the batch-1 shape.
scripts/check-dts-emitted.mjs	no named section banner in the self-test body (0 of the 2 the criterion needs): the body carries one prose comment and no ruled, boxed or enumerated section head.
scripts/check-logger-receiver-detach.mjs	one named section banner in the self-test body, below the two the criterion needs.
scripts/check-objectql-double-limit.mjs	one named section banner in the self-test body, below the two the criterion needs.
scripts/check-skill-compatibility-version.mjs	one named section banner in the self-test body, below the two the criterion needs.
scripts/check-skill-frame-freshness.mjs	no in-body assertion helper at all: cases are pushed inline at ~14 separate cases.push({ ... }) sites, so there is no single block body for registerCase(). The only block-bodied helper, linear(), is a fixture builder.
scripts/check-spec-parsed-alias.mjs	no named section banner in the self-test body: its comments label individual cases (// GOOD:, // BAD (rule 1)), not sections.
scripts/check-stack-collection-maps.mjs	no named section banner in the self-test body.
scripts/check-startup-registry-verdict.mjs	one named section banner in the self-test body, below the two the criterion needs.
scripts/check-tenant-chokepoint.mjs	no named section banner in the self-test body: three unmarked prose comments head its blocks, and reading one as a section head rather than an explanation of the next statement is a judgement per comment, which this transplant does not make.
scripts/check-turbo-task-graph.mjs	no body-level verdict site: selfTest() returns a failures array and the verdict is printed by a SEPARATE runSelfTest() (L892). That is the class-3 shape.
scripts/check-whole-set-label-write.mjs	one named section banner in the self-test body, below the two the criterion needs.
scripts/check-wildcard-fallthrough.mjs	no named section banner in the self-test body: four unmarked prose comments head its blocks.
scripts/js-comment-mask.mjs	no named section banner in the self-test body.
scripts/publish-smoke-pack.mjs	no comment at statement position in the self-test body at all, so no section banner.

Shrink-only ledgers: scanned, none found in this batch

Every battery that iterates a module-level constant was enumerated (28 battery/ledger pairs across 13 files) and each underlying list read: CATEGORIES, SHAPES, MUTATIONS, ROOTS, TEXT_STAMPS, EXPECTED_TEXT_FAMILY, FLOORS, REQUIRED_CONTEXTS, INSTRUCTION_SURFACES, RETIRED_CONTEXT_NAMES, EXPRESS_RESPONSE_MODULES, DECLARED_COUNTS, UNENFORCED_TEXT_COUNTS, ROOT_DIR_WATCH_HINTS, YAML_CENSUS_ANCHORS. None is shrink-only, and the two that read closest to it are not one-case-per-row: UNENFORCED_TEXT_COUNTS is asserted structurally (two cases, with the row count inside the case rather than as case multiplicity), and RETIRED_CONTEXT_NAMES only ever grows. So no floor in this batch moves with a list a legitimate edit shortens.

Evidence

Behaviour is unchanged, per file

For all 30: node FILE --self-test before and after, exit code captured by redirect before any pipe. 30/30 exit 0 both sides, and cmp -s reports stdout and stderr byte-identical on every one. (Both readings are from the tree after pnpm install; without it seven files exit 3 on a missing-dist prerequisite and two on ERR_MODULE_NOT_FOUND, which is a container state, not a verdict.)

Census

node scripts/measure-self-test-floor.mjs --json:

	merge base 2263ca4d6	this branch f248bc953
ROSTER	3	33
COUNT	4	4
NONE	158	128

Three set-equality assertions, computed in python, all True:

1. every one of the 30 worklist files reads NONE before and ROSTER after;
2. the set of files whose class changed equals the worklist exactly (symmetric difference empty);
3. git diff --name-only equals the worklist exactly.

And the 128 still NONE decompose exactly — a disjoint partition whose union is the set, asserted in python, not read by eye:

128 = 40 PR #14851 (batch 1, unmerged)
    + 18 PR #14872 (batch 1b, unmerged)
    + 23 PR #14853 (issue #13798 batch 2, unmerged)
    + 15 this batch's measured drops
    +  2 honoured i18n exclusions (PR #14856)
    + 13 class 2 — fewer than two banners, re-derived from the tree
    +  5 class 3 — no body-level verdict site
    +  5 Tier C / shape-blocked
    +  3 honoured open-PR exclusions (check-type-check-coverage, pm/dispatch-gates, pm/bare-root-worklist)
    +  4 outside the declared Tier-B population (check-platform-checklist, check-regen-pending, git-merge-regen, setup-git-hooks)

The 30 floored here intersect PR #14851's 40, PR #14872's 18, PR #14853's 24, the i18n pair and the open-PR exclusions in zero files.

Ablations — four, each from the committed implementation

Each leg carries a trap restore on absolute paths; the mutation is proved on disk by an anchored grep count plus git hash-object against the HEAD blob, and the restore by an empty git diff HEAD — never by an exit code. No dist is on the resolution path of any of the four, so no rebuild leg applies to them.

1. A battery's cases deleted — scripts/check-auth-mount-ledger.mjs, the 2 ok(deriveBasePath( cases (grep 2 → 0; cd7b24ab → b8493fe3). Exit 0 → 1: battery "The base path is DERIVED, and its absence is not an empty population." DID NOT RUN — 0 cases registered, 2 pinned. The verdict below would have claimed those cases hold.
2. A roster entry deleted — scripts/check-engine-split-ratio.mjs (grep 1 → 0; d92d2186 → 06deaef8). Exit 1 naming both halves: SELF_TEST_BATTERIES declares 1 batteries, below the pinned 2 and battery "pure rendering" registered 5 case(s) but is not declared in SELF_TEST_BATTERIES.
3. The introduced registerCase() removed from the sink helper — scripts/check-quick-reference-counts.mjs (grep 1 → 0; ce2b2de3 → a29feeba). Exit 1 with every original case still passing: all 22 batteries report DID NOT RUN.
4. One battery() opener deleted — scripts/docs-audit/affected-docs.mjs (grep 1 → 0; 965c95f0 → 255698df). Exit 1: battery "anchor PROVENANCE ([decision] docs-audit: a data-property anchor is both the noisiest and the most valuable anchor the tool mints — 70 of 402 rows, and no cheap discriminator survives measurement #12824)" DID NOT RUN — 0 cases registered, 12 pinned.

A fifth attempt aborted before running — the anchor matched zero lines and the guard refused rather than reporting a clean measurement, which is the failure shape the ablation contract exists to catch.

Gates

Derived with node scripts/pm/dispatch-gates.mjs --repo objectstack-ai/objectstack (no paths passed; the script reads the change set from git — 30 paths vs merge base 2263ca4d6). 55 commands derived + check:declared-population-live + check:pm-dispatch-gates + check:nul-bytes = 58 run. Exit codes captured by redirect before any pipe.

• 57 RAN-PASS (exit 0) of the 58 run, including both mandatory gates: check:ratchet-remedy-authority 0, check:declared-population-live 0 (158 of 202 famil(ies) declare a path population, and every one of them reaches this tree's 8079 tracked file(s)).
• check:pm-dispatch-gates — exit 0, dispatch-gates self-test: 1241 cases pass. No compound-anchor red: this batch introduces no top-level function declaration at all, so the SELF_TEST_DECL anchor sees nothing new. The battery/registerCase closures are function-scoped and the three new module-level names are consts, which the anchor does not match.
• check-test-completeness — exit 3, NOT MEASURED by the gate's own instruction, not a red.
• Two first-pass readings that were container state, both cleared and re-run green: check-dev-prereqs exit 1 (The workspace is not built — 1 unmet precondition, not a list of problems) → 0 after pnpm exec turbo run build --filter=./packages/* --filter=./packages/*/* under scripts/pm/os-verify-lock.sh (VERDICT command-exit 0, held 6m32s); check-engine-split-ratio exit 2 (this clone is shallow … a ratio derived here would be real, plausible and WRONG) → 0 after git fetch --shallow-since=2026-05-29 origin main. Neither is a verdict about this diff, and neither was read as a pass before it was re-run.
• pnpm lint (whole repo, eslint . --no-inline-config) — exit 0 on f248bc953. No narrowing claimed.
• Control-byte scan over the 30 changed files (grep -naP) clean; check:nul-bytes exit 0.

The build left the tracked tree unchanged (git status --porcelain empty).

scripts/** only ⇒ skip-changeset.

---

Generated by Claude Code

[github-actions]

📓 Docs Drift Check

Nothing in this diff resolved to a documentable surface (no symbol, route or SDK anchor derived from 0 changed package(s)), so this run has no opinion about the docs.

What this run could not see

• 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.

Coarse fallback — 0 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 f116b8f8d1b43ec2bfd64fbf0ebb0cf3c301b1c8 → packageMentionDocs.