What problem does it solve? Turning a GitHub issue into a merged pull request involves many failure points: stale issue threads, duplicate PRs from other contributors, misdiagnosed root causes, regression tests that prove nothing, and premature claims of green CI. This Skill enforces an end-to-end discipline so every fix is validated, deduplicated, class-complete, and honestly reported. ## Core Features & Use Cases - Premise Validation: Reads the full issue thread and git history to confirm the reported bug is real and not intentional design before writing code. - Duplicate and Class-Level Sweeps: Searches existing PRs and sibling call sites so work is not duplicated and the whole bug class is fixed in one PR. - Sabotage-Tested Regression Coverage: Temporarily restores old behavior to prove the new regression test actually fails without the fix. - Honest CI Shepherding: Distinguishes diff-introduced failures from baseline or infrastructure flakes and reports only live-verified CI state. - Use Case: A maintainer says "Fix issue #123 and open a PR." The Skill reads the thread, sweeps for duplicates, reproduces the bug, writes a proven regression test, fixes sibling call sites, opens the PR immediately, and monitors CI until the issue is closed with a traceable comment. ## Quick Start Ask the agent to fix GitHub issue number 123 in the current repository and open a verified pull request.