Monitor CI

Set up event-driven CI monitoring using Claude Code's Monitor tool. Replaces GitHub email notifications with agent-actionable events.

When to use

• At session start (captain sessions — monitor main branch health)
• After pushing a PR branch (watch for check results)
• When you want instant CI failure notification instead of checking email

Instructions

Start monitoring

Use the Monitor tool to run the ci-monitor script in the background:

Monitor CI for me. Run ./agency/tools/ci-monitor in the background persistently.
When output appears, investigate the failure and report.

The script:

• Polls GitHub Actions every 60 seconds (configurable with --interval N)
• Only outputs when there ARE failures (completely silent when green)
• Checks main branch workflow runs and open PR check statuses
• Output is prefixed with [CI] for routing

Options

./agency/tools/ci-monitor --interval 30          # Poll every 30 seconds
./agency/tools/ci-monitor --repo owner/repo      # Explicit repo (default: auto-detect)

When output arrives

1. Parse the failure information from the Monitor output
2. If main branch failure: investigate immediately — this is a broken window
3. If PR check failure: check if it's your PR, investigate if so
4. Report findings to the principal

Why this replaces email notifications

Aspect	Email	Monitor
Latency	Minutes to hours	60 seconds
Noise	Every failure, even expected	Only actionable failures
Action	Manual triage	Agent can investigate automatically
Signal	Learned to ignore	Always fresh

Stop monitoring

Ask Claude to stop the monitor, or it stops automatically when the session ends.

OFFENDERS WILL BE FED TO THE — CUTE — ATTACK KITTENS!