Monitoring

Monitoring is an on-demand request, not a mandatory step. The agent helps the user check their app's health or set up basic monitoring. Keep it simple.

If the user asks to check production health

Do not set up anything. Just check what is available:

1. Read .agents/skills/deploy-project/SKILL.md to find the deployment URL.
2. If a URL exists: fetch it and check for a successful response. Report status, response time, and any obvious errors.
3. If no URL: ask the user for the URL.
4. If the app is down or returning errors, report what you see. Do not fix anything unless asked.

If the user wants to set up monitoring

Ask: What do you want to monitor?

• Uptime (is the site alive?)
• Errors (are there failures in logs?)
• Performance (is it slow?)

Then ask: Do you already use a monitoring service? (Sentry, UptimeRobot, BetterStack, etc.)

• If yes: help them configure it for this project. Use the service's CLI or dashboard — never handle API keys directly.
• If no: suggest the simplest free option based on what they want:

For the simplest possible setup with no external service: suggest a health check endpoint and a cron job.

Health check endpoint

If the project does not have one, create a simple /health route that returns 200 OK. This is enough for basic uptime monitoring.

Simple cron monitoring (no external service)

Tell the user to add a cron job on their VPS or local machine:

# Check every 5 minutes, log failures
*/5 * * * * curl -sf https://yourapp.com/health || echo "$(date): DOWN" >> /var/log/monitor.log

For VPS deployments with Docker: add a lightweight monitoring container to docker-compose.yml only if the user wants it. Do not add it by default.

Save as project skill

After setting up monitoring, save to .agents/skills/monitor-project/SKILL.md:

---
name: monitor-project
description: Basic monitoring for this project.
---

# Monitor Project

- **URL**: https://yourapp.com
- **Health endpoint**: /health
- **Uptime check**: UptimeRobot (free) or cron job
- **How to check**: visit the URL or run `curl -sf https://yourapp.com/health`
- **Logs**: SSH to VPS, run `docker compose logs --tail 50`

Rules

• Never set up complex monitoring stacks (Prometheus + Grafana + Alertmanager) unless the user explicitly asks and understands the maintenance cost.
• Never handle API keys or tokens. Guide the user to add them in the service dashboard.
• If the user does not know what monitoring is, explain it in one sentence: "Monitoring tells you when your app is down or broken, so you can fix it before users notice."
• Keep it to one free tool maximum. Do not overwhelm with options.