What problem does it solve? Watching CI runs, paging through massive failure logs, and distinguishing real branch bugs from flakes or inherited main breakage consumes developer time and floods context with noise. This Skill automates that triage loop so the current branch's CI reaches green without manual log digging. ## Core Features & Use Cases - Current-head run detection: Finds the checks that actually cover the current HEAD SHA, waits on pending runs with a shared background poller, and never mistakes a stale green run for current coverage. - Context-safe diagnosis: Delegates large failing logs to a sweep subagent that returns failing jobs, decisive root-cause lines, and implicated code pointers instead of raw logs. - Report or fix modes: Reports the failing job, real error, cause, and concrete fix, or in fix mode applies the fix, pushes, and re-watches the new HEAD until green. - Use Case: After pushing a PR, ask the agent to check CI; it waits for the matrix to finish, identifies that a test failure is a branch bug rather than a flake, applies the fix, pushes, and confirms the new HEAD is green. ## Quick Start Run the ci skill in fix mode to diagnose this branch's failing CI and push fixes until the current HEAD is green.