Monitor & Alert

---
name: monitor
description: >
  Set up monitoring, alerting, and on-call coverage for a service.
  Wires up structured logs, error tracking, uptime checks, latency
  dashboards, and Slack/PagerDuty alert routing with sensible defaults.
  Use when launching a new service, hardening a side project before
  promotion, or closing the loop after an incident by adding the missing
  telemetry that would have caught it.
when-to-use:
  - Wiring up a new service end-to-end on day one
  - Adding uptime and error-rate alerts before a launch
  - Closing the telemetry gap after an incident
  - Replacing a fragmented observability stack with one defaults file
---

# Monitor & Alert

Stand up a complete observability baseline in a single pass — logs,
errors, uptime, latency, and alerts — without writing glue code.

## Overview

The skill generates a single `monitor.config.yaml` plus the integration
snippets needed to cover four pillars:

1. **Logs** — JSON-structured logger config (Pino / winston / slog) plus
   a drain to your aggregator (Datadog, Better Stack, Grafana Loki).
2. **Errors** — Sentry init with sensible sampling, release tracking, and
   sourcemap upload wired into the deploy skill.
3. **Uptime** — three-region HTTP / TCP checks against `/healthz` and key
   API paths, with synthetic-browser checks for marketing pages.
4. **Alerts** — error-rate, p95 latency, and saturation rules with
   severity tiers, routing to Slack by default and PagerDuty for `sev1`.

It also produces a runbook template with the first five alert paths so
the on-call engineer has somewhere to start at 3 a.m.

## Usage examples

```bash
# Wire everything up for a Node service
npx skills run monitor --service api --provider vercel

# Add Sentry + uptime only, skip logs
npx skills run monitor --service api --include sentry,uptime

# Generate a runbook from existing alerts
npx skills run monitor --runbook-only --provider betterstack
```

## Parameters

| Name | Type | Default | Description |
|---|---|---|---|
| `service` | string | — | Logical service name; used as the Sentry project and the Slack channel prefix. |
| `provider` | enum | auto | `sentry`, `datadog`, `betterstack`, `grafana`, `newrelic`. |
| `include` | string[] | `all` | Subset of `logs`, `errors`, `uptime`, `alerts`, `runbook`. |
| `regions` | string[] | `iad,fra,syd` | Uptime check regions. |
| `routes` | string[] | `["/healthz"]` | Paths to probe. |
| `pagerduty_key` | string | — | Service key for `sev1` escalation. |
| `slack_channel` | string | `#oncall` | Default alert destination. |
| `runbook_only` | bool | `false` | Skip wiring and just generate the runbook markdown. |

## Expected output

`monitor.config.yaml`, integration code snippets, and a `RUNBOOK.md` with
one section per alert rule: trigger condition, blast radius, first three
diagnostic steps, and the rollback command. The skill exits non-zero if
any provider auth is missing so CI fails loudly.