What problem does it solve? When exactly one CI check on a pull request turns red while everything else is green, it is ambiguous whether the failure is a real finding in your diff, pre-existing debt, or a transient infrastructure issue. This Skill settles the question with local evidence before you fix, re-run, or escalate. ## Core Features & Use Cases - Local gate reproduction: Maps the failing check to its underlying tool and runs the full local equivalent (lint, unit, E2E, mutation targets) rather than a narrowed invocation. - Finding attribution: Attributes every finding to a file and compares it against the diff to classify the failure as environmental, a real regression, or pre-existing debt. - Escalation path: Defines when to hand off to infrastructure-failure diagnosis, with the evidence (commands, exit codes, runner logs) needed for an actionable re-run request or issue. - Use Case: A Playwright shard fails on your PR with a timeout while all other checks pass. Run the local E2E target, confirm a clean result, and record the evidence in the PR thread to justify a re-run. ## Quick Start Ask the assistant to validate why the single failing CI check on my pull request is red while all other checks are green.