Summary
The #291-#294 patch set wires the test pyramid more cleanly, but the repository still lacks an enforceable, transparent coverage policy that maps to the desired unit/integration/e2e split.
Evidence
- The current CI lanes are separated (
unit, integration, e2e-smoke, e2e-full, fuzz), but the repo does not yet have a single accepted coverage contract for these lanes.
- Targeted package coverage after the patch remains uneven across large packages, so a generic “90%+” claim would be misleading without agreed measurement rules.
Why this matters
- Teams will optimize whatever is measured. Right now lane wiring exists, but the coverage target is not operationalized.
- Without explicit budgets, future contributors can unintentionally shift coverage away from the intended test pyramid.
Proposed work
- Define the measurement scope: repo-wide, changed-package, or changed-line coverage.
- Publish budgets for each lane, for example unit as the primary coverage source with integration/e2e focused on contract and workflow confidence.
- Emit machine-readable coverage summaries per lane and fail CI when agreed thresholds regress.
- Add documentation so local
npm run and CI use the same definitions.
Acceptance criteria
- Coverage policy is documented in-repo.
- CI reports lane-specific coverage metrics in a stable format.
- Threshold regressions fail predictably and do not depend on local tool quirks.
Summary
The #291-#294 patch set wires the test pyramid more cleanly, but the repository still lacks an enforceable, transparent coverage policy that maps to the desired unit/integration/e2e split.
Evidence
unit,integration,e2e-smoke,e2e-full,fuzz), but the repo does not yet have a single accepted coverage contract for these lanes.Why this matters
Proposed work
npm runand CI use the same definitions.Acceptance criteria