- •Read
codex.toml(if present) for canonical commands; otherwise use sensible defaults. - •Run in this order (skip steps not applicable to the repo):
- •format/lint
- •typecheck
- •unit tests
- •feature-scoped tests (by path/marker if available)
- •Summarize results:
- •What passed
- •What failed (top 3 failure signatures)
- •Likely category: env/setup vs flaky vs deterministic bug vs expectation mismatch
- •Recommend the next action:
- •
debug-loopfor deterministic failures - •
failure-triageif there are many failures - •
flaky-test-hunterif it appears intermittent
- •
test-runner
持续执行正确的检查,并将失败情况归纳为可操作的类别。
SKILL.md
--- frontmattername: test-runner description: Run the right checks consistently and summarize failures into actionable buckets.