What problem does it solve? A check that never ran, ran on the wrong object, or ran on trivially passing input reports exactly the same green status as a genuinely passing check. This Skill helps you detect and fix gates, tests, and CI steps whose "clean" result means "nothing was actually examined" — before the bug they guard ships anyway. ## Core Features & Use Cases - Five measured failure mechanisms: scope excludes the guarded case (disabled objects), malformed input makes the check succeed trivially, errors land on an unread channel (stderr vs stdout), plumbing defaults silently kill an input (shallow clones), and a guard fires but fixing forward leaves the damage in place. - Two-clause verification test: make the check fail on purpose once, and check what the wrong state already cost before clearing it. - Review checklist and baselining guidance: a checklist for new or inherited checks, plus how to baseline a gate that ships red so it fails only on new divergence with verified (not plausible) reasons. - Use Case: Before trusting a release workflow, a CI gate, or a test you have never watched fail, walk the checklist — e.g., verify a paths: ['VERSION'] release filter actually triggers, and confirm via gh release view that the release exists rather than trusting the green tick. ## Quick Start Ask the assistant to audit a specific CI gate or test using the green-gate checklist and make it fail on purpose once to prove it can fail.
