instant-observability

Install

$ curl -fsSL -L https://tetsukod.ai/api/skills/instant-observability/download | tar -xz -C ~/.claude/skills/

Drops the `instant-observability/` skill into `~/.agent/skills/`. Once it lands, `/instant-observability` is available in any Claude Code session run from a project root.

What it does

Stand up a production-shaped logging pipeline for a Node project in one interactive run — Loki for storage, Grafana for query, Pino for the application logger, structured logs end-to-end. The skill provisions the stack on the deployment target you pick, wires the logger into your codebase, drops the Grafana MCP into your project so future Claude Code sessions can query the logs directly, then refactors the codebase off raw console.* calls onto a single structured logger with sensible levels.

What you get:

• A working Loki + Grafana stack on local Docker, a remote server via SSH, or Grafana Cloud — your call, asked once at the start
• A single Pino logger module wired to ship to Loki with structured fields, env-driven config, and a redact block for secrets
• The Grafana MCP server in your project's .mcp.json — future agents query logs through it instead of docker compose logs
• A console.* to logger.<level> refactor pass that classifies trash vs signal per a bundled rule sheet and pauses for diff approval per file
• New instrumentation at request boundaries, external-API call sites, and catch blocks that previously swallowed errors

What's inside

The package ships with every template the skill installs into your project — a Loki + Grafana docker-compose, a preprovisioned Loki datasource for Grafana, the canonical Pino logger module with a pino-loki transport, the .mcp.json fragment for the Grafana MCP server, and the trash-vs-signal decision sheet the refactor pass reads. The orchestrator in SKILL.md runs the interview, the smoke test, and the per-file refactor gate.

When to reach for it

Reach for this skill when a Node project is past the prototype stage and you want real log infrastructure today — not next sprint, not after a "let's pick a logging library" meeting. It is the right fit for greenfield projects, projects buried under accumulated console.log breadcrumbs, and projects where you want future Claude Code sessions to query logs directly through the Grafana MCP. Once Grafana is up, run /instant-mobile-notifications next to route alert rules to your phone as ntfy push notifications — the alerting axis is decoupled from the log pipeline on purpose, so each skill stays small enough to run cleanly in one shot.

Skip it for non-Node codebases — the bundled Pino module is JS/TS-shaped. Skip it for production hardening that needs S3-backed Loki storage and multi-tenant auth, since this is the instant-on shape. And skip it for tracing or metrics, which are separate skills not yet authored.