chore(scripts): bounded audit of bookkeeping edits forced on feat and fix PRs

[thymikee]

Child 8 of #2409.

Purpose

Find out whether any bookkeeping under scripts/ still forces edits on ordinary feature and fix PRs, and remove only what replays as a saved file without weakening independent evidence. Bounded: close with no implementation if nothing qualifies.

Baseline

Measured on main at e7d97f7dfe, recent 456 code commits:

• 24% of feat+fix commits touch scripts/; 8% touch scripts/layering/. Refactor commits: 67% and 61%. The layering churn is the migration's own cost and fell from 35 to 6 per 100 commits.
• feat+fix files under scripts/ with the most edits: scripts/layering/package-boundaries.test.ts 6, scripts/__tests__/eager-closure-budgets.ts 5, scripts/maestro-conformance/build-manifest.mjs 4, scripts/maestro-conformance/corpus/manifest.json 4, scripts/__tests__/test-file-size-ratchet.test.ts 4.
• Already ruled out: the eager-closure budget and the test-file-size ratchet compare against the merge-base and need a hand row only for approved exceptions; the package export snapshot in package-boundaries.test.ts:55 is an independently reviewed surface by design; Maestro fixtures are pinned upstream artifacts. Replaying the removal of every such edit over the recent 186 feat+fix commits moves the median files per commit from 8 to 8.

Required work

1. Take the ten most recent feat or fix commits that touched scripts/. For each scripts/ file edited, classify the edit: independent evidence (keep), approved exception row (keep), or duplicate bookkeeping that restates something the same PR already declares elsewhere.
2. For each duplicate, show the historical replay: the same PR with that edit removed and the check still red against a planted violation.
3. Remove only those. Post the table either way.

Acceptance

• The table of ten commits with a classification per scripts/ edit is posted on this issue.
• Each removal shows a replay of F → F−1 on a real historical commit and a red gate against a planted violation.
• No gate is replaced by self-comparison of the implementation under test.

Non-goals

• Touching the layering gates while the extraction program is still running.
• Regenerating Maestro conformance fixtures in ordinary tests.

[thymikee]

Bounded audit: ten most recent feat/fix commits touching scripts/

Commits (newest first), git log --no-merges on origin/main (head e7d97f7dfe), filtered to subjects starting with feat/fix and touching scripts/:

#	Commit	Subject
1	d11c8cf9d6	feat: support standalone Maestro clearState command (#2366)
2	3f022b0730	fix(gates): stop an eager-closure approval from turning main red on merge (#2375)
3	7bea29d61e	fix: repair main after the managed-allocation move landed under stale PRs (#2328)
4	ebdaa7617e	feat: delegate reviewed managed automation (#2312)
5	96727a0b42	fix(apple-runner): never compare an unavailable toolchain probe; name the mismatching cache keys (#2306)
6	fa06c8c6e9	fix(daemon): fence managed readiness behind runtime admission (#2280)
7	5bb3ea3b2a	feat(ios): productionize Simulator AX snapshot bridge (#2277)
8	172ee149cf	feat(screenshot): add --crop-on to crop captures to a selector frame (#2276)
9	658f822c40	fix: encode the mcp subcommand in server.json package arguments (#2275)
10	e882cf9723	feat(runtime): add managed-local ownership and the exact-only managed runtime (#2258)

Per-file classification

#	Commit	scripts/ file	Classification
1	d11c8cf	fuzz/validation-arbitraries-maestro.ts	independent evidence — new fuzz coverage for the clearState command shape
1	d11c8cf	maestro-conformance/build-manifest.mjs	approved exception row — generator's NOTES map for the new authored corpus flow
1	d11c8cf	maestro-conformance/corpus/authored/clear-state.yaml	approved exception row — pinned corpus fixture
1	d11c8cf	maestro-conformance/corpus/manifest.json	approved exception row — generated manifest entry paired with the build-manifest.mjs NOTES row above (generator + checked-in output, not a redundant check)
1	d11c8cf	fixtures/layer1-parser.json	approved exception row — pinned upstream parser fixture
2	3f022b0	__tests__/committed-source-tree.ts	independent evidence — new headCommit helper backing the merge-base-is-head fix
2	3f022b0	__tests__/eager-closure-budgets.test.ts	independent evidence — coverage for the new staleApprovalRows rule
2	3f022b0	__tests__/eager-closure-budgets.ts	independent evidence — genuine mechanism fix (merge-base-vs-head bug), not bookkeeping
3	7bea29d	__tests__/eager-closure-budgets.ts	approved exception row — stale APPROVED_OVER_CEILING row removed exactly as the rule prescribes
4	ebdaa76	layering/package-boundaries.test.ts	approved exception row — independently reviewed export-snapshot surface (established)
5	96727a0	write-xcuitest-cache-metadata.mjs	independent evidence — production CI script logic fix (fail-closed toolchain probe)
6	fa06c8c	layering/architecture-ownership.ts	approved exception row — same archetype as the package-export snapshot: architecture-ownership.test.ts asserts the façade's actual named exports against this hand-declared list, so it is an independently reviewed surface, not a restatement of anything else in the PR
7	5bb3ea3	__tests__/fixtures/size-report-npm-pack.json	independent evidence — fixture data for new test cases in the same PR
7	5bb3ea3	__tests__/size-report-package.test.ts	independent evidence — new assertions for assertPublishPackageContents's new requireSnapshotBridge behavior
7	5bb3ea3	check-package.ts	independent evidence — separate install path (installIntoCleanConsumer) exercising the same invariant size-report-install.mjs exercises on its own (benchmark) install path; distinct call sites, not a restatement
7	5bb3ea3	ios-snapshot-benchmark/size-install.test.ts	independent evidence — new coverage for assertInstalledSnapshotBridge
7	5bb3ea3	layering/package-boundaries.test.ts	approved exception row — export-snapshot surface (established)
7	5bb3ea3	layering/platform-package-policy.test.ts	approved exception row — same hand-declared-surface archetype as architecture-ownership, applied to the new snapshot-source subpath
7	5bb3ea3	layering/platform-package-policy.ts	approved exception row — declares the new snapshot-source subpath the policy test above checks
7	5bb3ea3	size-report-install.mjs	independent evidence — post-install-directory check; distinct failure mode from the packed-tarball check below (catches install-time/publishing divergence, not packaging-config divergence)
7	5bb3ea3	size-report-package.mjs	independent evidence — packed-tarball-listing check (assertSnapshotBridgeAssets), distinct from the post-install check above
8	172ee14	integration-progress-model.ts	approved exception row — new named exemption row (screenshotCropOn) with owner + reason, same pattern as APPROVED_OVER_CEILING
8	172ee14	layering/package-boundaries.test.ts	approved exception row — export-snapshot surface (established)
9	658f822	sync-mcp-metadata.mjs	independent evidence — production sync-script logic fix (adds the missing mcp positional argument)
10	e882cf9	layering/package-boundaries.test.ts	approved exception row — export-snapshot surface (established)

Result

No file edit across the ten commits classifies as duplicate bookkeeping. Every scripts/ edit is either:

• independent evidence (a real behavior/test change backing the PR's own fix or feature), or
• an approved exception row (a hand-maintained, independently-checked declaration — merge-base-relative budget rows, the package/façade export-surface snapshots, or a named test-exclusion row with owner and reason).

No removal is proposed. Per the issue's bounded-audit acceptance criteria, closing with no implementation.

[thymikee]

Audit complete: the ten-commit classification table is posted above and found no duplicate bookkeeping — every scripts/ edit in feat/fix PRs is independent evidence or an approved exception row. Closing with no implementation, as the issue specified. Part of #2409 (child 8: forecast 0, confirmed).