Debug GitHub Actions Failures with Claude

Failed CI run with 500 lines of logs? Don’t scroll through it. Feed it to Claude.

Pipe Logs Directly

gh run view --log-failed | claude -p "Why did this CI run fail? Give me the root cause and a fix."

Specific Workflow Run

# List recent failed runs
gh run list --status failure --limit 5

# Get logs for a specific run
gh run view 12345 --log-failed | claude -p "Diagnose this CI failure."

In-Session Debugging

Run `gh run view --log-failed` and tell me:
1. Which step failed
2. The root cause
3. How to fix it

Claude runs the command, reads the output, and gives you a diagnosis.

Common Patterns Claude Catches

• Dependency version mismatches between local and CI
• Missing environment variables or secrets
• Flaky tests that pass locally but fail in CI
• Docker layer caching issues
• Node/Python version differences

Tip

Add this to your CLAUDE.md so Claude always uses gh for GitHub operations:

Use the `gh` CLI for all GitHub operations (issues, PRs, actions, runs).
Don't try to use the GitHub API directly.