Agent Regression Check

Evaluate whether an agent change introduced regressions.

Compare matched before vs after case results, score degradation with a deterministic rubric, apply release gates, and return one verdict:

go
conditional_go
no_go
rollback
This is an offline regression-check skill. It does not replace
production monitoring, live experiments, or telemetry.

When to Use
Use this skill when the user asks:

Did this update break anything