canarylisted

# Canary I'll watch your deployment after it goes live, polling for errors and regressions until health is confirmed or a problem surfaces. Inspired by gstack's canary skill. ## Token Optimization **Expected range**: 200–600 tokens per poll cycle, 50 tokens (healthy exit) **Patterns used**: Bash for all checks, early exit (all green on first check), progressive disclosure (status line per cycle → details only on failure) **Early exit**: If all checks pass on the first cycle, report "Deployment healthy — all checks green" and stop. ## Step 1 — Identify What to Monitor Infer from the project or accept explicit arguments: ```bash # Check for common deployment indicators ls -la .env .env.production docker-compose.yml Procfile 2>/dev/null # Check for health endpoint conventions grep -r "health\|ping\|status" --include="*.json" --include="*.yaml" -l 2>/dev/null | head -5 # Check for error log locations ls -la logs/ /var/log/app.log 2>/dev/null ``` If no targets are auto-detected, ask: - What URL or endpoint should I poll? - Where are the application logs? - What error patterns should I watch for? ## Step 2 — Configure the Watch Loop **Defaults (override via arguments):** - Poll interval: 30 seconds - Max duration: 10 minutes - Error threshold: 2 consecutive failures = alert - Success threshold: 3 consecutive passes = declare healthy ```bash # Example: /canary --url https://api.example.com/health --interval 30 --duration 10m ``` ## Step 3 — Poll Loop Each cycle runs