FixCI Workflow
Steps
1. Verify Authentication
gh auth status
If not authenticated, stop and ask user to run gh auth login.
2. Resolve PR
If user provided a PR number or URL, use it. Otherwise resolve from current branch:
gh pr view --json number,url
3. Inspect Failing Checks
Run the bundled script first:
python3 "<skill-dir>/scripts/inspect_pr_checks.py" --repo "." --pr "<number>"
The script:
- Fetches all PR checks
- Filters to failing ones (failure, cancelled, timed_out, action_required)
- Pulls GitHub Actions logs for each failure
- Extracts failure snippets around error markers
- Reports run metadata (workflow, branch, SHA)
4. Analyze Failures
For each failing check:
- Read the failure snippet from script output
- Identify the root cause category:
- Test failure: Which test(s) failed, what assertion broke
- Build failure: Compilation error, missing dependency
- Lint/format: Style violation, formatting issue
- Timeout: Which step hung, resource exhaustion
- Infra: Runner issue, network failure, flaky setup
- Check if the failure is actionable (code fix needed) vs transient (retry may fix)
For GitHub Actions-specific issues (workflow config, permissions, caching), delegate to reviewer-ghactions subagent for deeper analysis.
5. Summarize and Plan
Present a summary:
## CI Failures for PR #<number>
### <Check Name> — <Category>
**Status:** <conclusion>
**Workflow:** <workflow name>
**Cause:** <one-line explanation>
**Evidence:**
…
### 6. Wait for Approval
**CRITICAL: Do NOT edit code without explicit user approval.**
Ask: "Should I implement these fixes?" or present options if multiple approaches exist.
### 7. Implement Fix
After approval:
1. Make the code changes
2. Run relevant local tests to verify
3. Report what was changed
### 8. Verify
After fix is pushed:
```bash
gh pr checks <number> --watch
Or suggest user to check CI status after push.