handle-ci-failurelisted

# Handle CI failure ## Project CI context ```! cat .github/HANDLE_CI_FAILURE.md 2>/dev/null || echo "No .github/HANDLE_CI_FAILURE.md found. Derive the branch from \`git branch --show-current\`; find the check command from the Makefile, CI workflow, or package scripts (or ask the user); push without force unless you confirm the branch is rebased on a cycle. There are no project-specific failure patterns to consult." ``` The context file is the project's contract with this skill. It declares the working branch, the check command, the push policy, the commands behind any auto-fixable failures (formatter, regenerate, dependency tidy), and any project-specific failure patterns. Read it before matching. ## Inputs The user may provide: - A bare run ID (`24375242526`). - A `View results:` URL from a GitHub notification email — extract the numeric run ID with `grep -oE 'actions/runs/[0-9]+' <<<"$input" | cut -d/ -f3`. - Nothing — list recent failures on the working branch and ask. ## Workflow ### 1. Resolve the run ID If the user gave one, use it. Otherwise: ```bash gh run list --branch <branch> --status=failure --limit 5 ``` Show the list with run IDs and ages, and ask which one to handle. Don't silently pick "most recent" — it drifts between sessions. ### 2. Extract failure context ```bash "$CLAUDE_SKILL_DIR"/scripts/ci-failure-context.sh <run-id> ``` Output JSON shape: ```json { "run_id": "...", "run_url": "...", "workflow": "...", "branch": "...", "head_sha":