Baseline debt — standing report

[github-actions]

Baseline debt — 17 tracked

Stalest first. No deadlines: these are measurements, not commitments.

Baseline	Entries	Unchanged for	Target
src/utils/markdownPipeline/__tests__/fidelity/fidelityLedger.ts	—	36d	down — same
scripts/plugin-store-coupling-baseline.json	23	34d	zero on every channel — options and seams, not ambient lookups
src/utils/markdownPipeline/conformance/expectedDeltas.ts	—	33d	down — same
src/utils/markdownPipeline/__tests__/spec/specDeltas.json	—	22d	down — each declared divergence fixed or ruled a deliberate dialect choice
src/utils/markdownPipeline/__tests__/spec/specRoundtripDeltas.json	—	22d	down — same
scripts/command-error-baseline.json	23	21d	zero — every command returns Result<T, CommandError>
scripts/theme-contrast-baseline.json	6	8d	failing lists empty for every theme — Phase 1 of the UI-consistency plan retints the catalog; ansiFloor/exempt stay as reasoned permanent exceptions (D10)
scripts/ui-consistency-baseline.json	6	8d	every list empty — C10 empties in WI-UI1.8, C5 in WI-UI2.1, C7/C8 in WI-UI2.3, C4 in WI-UI3.1/3.2, C9/C11 in WI-UI3.x; permanent exceptions become ui-ok markers, not entries
scripts/gate-tests-baseline.json	9	8d	empty — each listed gate gains a sibling scripts/.test.* and its entry is removed
scripts/i18n-copy-baseline.json	40	8d	empty — each remaining multi-sentence description is reworded (or Q3 is revisited for multi-sentence copy) and its entry removed
scripts/bespoke-buttons-baseline.json	156	5d	down — each bespoke class converted to .vm-btn / .popup-icon-btn
scripts/design-tokens-baseline.json	9	5d	both lists empty — each rgba literal moves onto a token (or the sanctioned color-mix pair) and each className literal onto a token-backed class (WI-UI1.x/2.2)
scripts/mock-boundaries-baseline.json	265	5d	zero — tests use the real store (setState/reset) or a recorded store-factory seam
.dependency-cruiser-known-violations.json	74	4d	down — each frozen edge either deleted or re-expressed through a plugins/shared seam
scripts/extension-budget.json	5	4d	down — per-edge licenses retired as plugins decouple
scripts/file-size-baseline.json	157	0d	down — split files, lower the number; never raise
scripts/test-types-baseline.json	349	0d	zero — every test file typechecks under tsconfig.test.json; 2,503 errors in 376 files on adoption, all pre-existing

A row that never moves is the finding. Entries counts the records the ratchet
itself compares — not a summed magnitude, because the values mean different
things per baseline (file-size holds line counts, command-error holds signature
counts). — means a custom comparator owns the shape and no honest total exists.

Filed by baseline-review — run 34099049320.

gate-liveness watcher

  ✓ gate-liveness.yml: last verdict 4.0d ago (cadence 8d)

✅ all 1 liveness gate(s) are producing verdicts

[xiaolai]

Correction on this report's numbers, for anyone reading it later: the "Unchanged for" column was wrong — every row read 0d because of how the job checked out the repo, not because the baselines had all just moved.

actions/checkout defaults to fetch-depth: 1, so the git log -1 -- <path> behind that column could only ever see the single fetched commit and returned the run's own timestamp for every baseline. Measured on a full clone at the time this was filed:

Baseline	Reported here	Actual
fidelityLedger.ts	0d	8d
.dependency-cruiser-known-violations.json	0d	6d
plugin-store-coupling-baseline.json	0d	6d
extension-budget.json	0d	6d
specDeltas.json / expectedDeltas.ts / specRoundtripDeltas.json	0d	5d
mock-boundaries / bespoke-buttons / command-error	0d	1d

The Entries counts were correct throughout; only the ages were not. The stalest-first ordering was therefore also arbitrary.

Fixed in #1251: the job now fetches full history, and --report refuses to publish rather than printing a table of zeros if it ever finds itself in a shallow clone again — a table of zeros is indistinguishable from a healthy register, which is why this went unnoticed. The next scheduled run will update the body above with real ages.

Substantively the register looks healthy rather than stalled: every tracked baseline had moved within 8 days.