[quality] vitest coverage thresholds sit ~11 points below actual coverage; the gate cannot catch regressions #201

Finding

vitest.config.mjs enforces coverage thresholds of lines 85 / branches 83 / functions 91 / statements 85. Actual coverage on main @ 187c5e3 (npx vitest run --coverage, 1291 tests) is:

A PR could delete roughly a quarter of the suite's coverage and build-test would still pass. The threshold exists as a regression gate but is not currently gating anything. There is no other coverage signal: .github/workflows/test.yml runs npm run test:coverage with text reporters only — no artifact, badge, or PR annotation — and the repository has no end-to-end suite, so unit coverage is the only evidence available.

Recommendation

Ratchet the thresholds to just under the current floor so a regression of more than ~1 point fails CI:

thresholds: {
  lines: 95,
  branches: 93,
  functions: 98,
  statements: 95,
},

Verified locally: the suite passes at these values on 187c5e3. Once the currently held error-path test PRs (#173–#200) land, coverage will be ~99% and the thresholds can be ratcheted again; alternatively set coverage.thresholds.autoUpdate: true so vitest rewrites the config whenever coverage rises, at the cost of a config diff in contributor working trees.

Priority

• Impact: medium — coverage regressions currently pass CI unnoticed
• Effort: low — one config change

Filed by quality agent (hold-gated mode)

🐝 Hive Agent: quality | Instance: hosted-available-lke648397-260827-5q9t | SHA: 187c5e3

— hive: agent=quality backend=copilot model=claude-fable-5.1 copilot=1.0.88