Regression Manage
Doctrine (v0.1.50): 把修法寫進工具,而非寫進 prompt. Programs first; AI is the backstop for owner-assignment narrative.
Once verification, synthesis, P&R, and signoff are scripted, the next bottleneck is keeping the regression green. This skill triages nightly / CI results and routes failures to the right skill / owner.
When to use
- After each nightly regression completes
- When a critical failure appears mid-day
- When flaky tests are slowing down the team
- Weekly regression health review
Inputs
- Regression run log / database (Jenkins, LSF, Slurm, GitHub Actions)
- Known-failure list with owners
- Severity policy (tape-out blocker vs warning vs info)
- Branch / tag under test
Technical basis
Test-farm orchestration patterns from large-scale SoC projects. Flaky-test management best practices from Google / Meta engineering blogs. LSF / Slurm / Jenkins are the standard back-ends.
Handoff
- P0 failures → direct routing to failing-step skill
- Repeated flaky →
/testbench-genor/coverage-closure(for stimulus stability) - Trending risk →
/tapeout-checklistas a red flag