| name | corgi |
| description | Author and explain corgi-compose.yml files and the corgi CLI. Use when writing or editing a corgi-compose.yml, picking db drivers / healthchecks / tunnels, or explaining corgi concepts, commands, and flags. Corgi is a Go CLI (`brew install andriiklymiuk/homebrew-tools/corgi`) that spins up databases, services, and required tools from one yml file — think docker-compose for services plus databases plus tool checks. NOT for starting a stack (use the run skill) or diagnosing a broken one (use the debug skill). Use when this capability is needed. |
| metadata | {"author":"Andriiklymiuk"} |
Corgi
Corgi (https://github.com/Andriiklymiuk/corgi) runs multi-service projects from a single corgi-compose.yml. It handles: cloning service repos, starting databases in Docker, seeding from dumps, generating .env files with cross-service URLs, and running all services concurrently.
When this skill activates, you are the expert on corgi. Do not fall back to generic docker-compose / npm advice if a corgi-compose.yml is present — corgi is the authoritative entry point for that project.
Critical: corgi run is long-running
corgi run blocks indefinitely and streams logs. Never run it synchronously — it will hang your shell. For agents, prefer corgi run --detach (returns immediately; see the lifecycle line below). See references/long-running.md before invoking it foreground.
Safe synchronous probes:
corgi doctor (alias check) — preflight: tools installed, Docker up, ports freecorgi status (aliases health, healthcheck) — post-run: TCP/HTTP probe each port
Both exit 0 on success, 1 on failure. Add global --json for machine-readable output. Driving corgi from an agent: see ../../../docs/agents.md.
Driving corgi as an agent
Corgi auto-detects non-interactive mode (CI/agent env vars like CLAUDECODE, or no TTY) and errors with exit code 2 instead of prompting; --interactive forces prompts back. Global --json emits pure JSON on stdout (doctor, status, list, config, ps, docs --json-schema; run --json prints a startup summary then streams logs to stderr). Exit codes: 0 success, 1 failure, 2 usage/missing input. Full guide + recipes: ../../../docs/agents.md.
Detached lifecycle (preferred for agents): corgi run --detach (starts services that outlive corgi, returns immediately, errors ALREADY_RUNNING if already running — use --force), then corgi ps/status for real running/crashed status, corgi stop [--service x] to tear down, corgi restart for full-stack restart. See the "Lifecycle (detached)" section in ../../../docs/agents.md.
corgi tunnel is also long-running (one tunnel subprocess per service, blocks until Ctrl+C). Background it the same way you background corgi run. See references/long-running.md if invoking from an agent.
Routing: when to read what
| Task | Read |
|---|
Writing a new corgi-compose.yml | references/yml-schema.md then references/db-drivers.md |
| Picking a db driver (port, image, env prefix) | references/db-drivers.md |
Adding healthCheck: to a service or db | references/healthchecks.md |
corgi doctor or corgi run failed | references/debugging.md |
| Explaining / choosing a CLI flag | references/commands.md |
| Setting up webhook tunnels (Stripe/GitHub/e-sign/etc.) | ../../../docs/tunnel.md (full) or references/commands.md#corgi-tunnel-services |
| Producing a service map / relation diagram for the project | run /corgi-describe (see references/describe-output.md) |
Running corgi run inside an agent loop | references/long-running.md |
Driving corgi non-interactively (--json, exit codes, scaffolding flags) | ../../../docs/agents.md |
| Driving corgi from an MCP client (Claude Code/Desktop tools) | ../../../docs/mcp.md |
| Persisting / re-reading service logs after a crash | references/commands.md#corgi-logs (corgi run --logs + corgi logs) |
| Opening a DB shell (psql / redis-cli / mongosh / …) | references/commands.md#corgi-db-shell-service-name |
| Running corgi in CI / disabling spinners and color | references/commands.md#corgi-run-flags (--ci, auto-detected from CI env) |
| Desktop notifications on service crash | references/commands.md#corgi-notifications (corgi notifications on|off|test) |
Load only what the task needs. Do not read every reference every time.
Common workflows
Fresh project from scratch: use the /corgi-new slash command, or: write corgi-compose.yml → corgi doctor → start corgi run in background → corgi status.
Document an existing project: /corgi-describe parses corgi-compose.yml and writes a detailed Markdown doc (services, dbs, env wiring, tunnels, scripts) plus a Mermaid relationship diagram to docs/corgi-services.md. Read-only — does not touch services. Built-in corgi --describe only prints per-service JSON during parse (and does not short-circuit — the underlying command, e.g. run, still executes); the slash command is the richer, side-effect-free alternative.
Existing repo with corgi-compose.yml: this file is the single source of truth for how services start. Do not invent npm run dev, docker compose up, or per-service shell commands — use corgi run. Look at db_services: to know what databases exist and at services: to know what service repos are expected.
User says "start the project" / "run the backend": check for corgi-compose.yml first (ls corgi-compose.yml or ls *.corgi-compose.yml). If present, corgi is the answer.
Quick command cheatsheet
corgi run # start everything (long-running, background it)
corgi run --logs # also persist each service's stdout/stderr to corgi_services/.logs/
corgi run --ci # CI-friendly: suppress spinners, banners, colors (auto-on when CI=true)
corgi run --service-branch api=feat/x # run a service on a branch in an isolated reused worktree (non-destructive)
corgi run --service-dir api=/path # run a service from an existing dir (e.g. a git worktree); repeatable, mix freely
corgi worktree list|prune # list / remove worktrees corgi made for --service-branch
corgi logs # interactive picker: browse + tail persisted logs (needs prior --logs run)
corgi logs --service api # skip service picker, jump to run picker for "api"
corgi logs --all # merge newest run of every service into one timestamp-sorted stream
corgi logs --idle 0 # tail forever (default exits after 30s of dead-air)
corgi logs --prune # delete all captured logs
corgi doctor # preflight checks (tools, docker, ports)
corgi validate # statically validate corgi-compose.yml
corgi test --service api # run a service's `test` script in its resolved env (--ensure-deps waits on deps)
corgi exec api -- <cmd> # one-off command in a service's resolved env (migrate/seed/etc); --ensure-deps to wait on deps
corgi env api # print a service's fully-resolved environment (read-only)
corgi config # show current user settings (~/.corgi/config.yml)
corgi notifications on|off|test # toggle/test desktop crash notifications
corgi status # health check (one-shot)
corgi status --ready # block until all healthy / timeout (CI-friendly)
corgi status --watch # live monitor, transitions only
corgi tunnel # public HTTPS tunnels (long-running) — default cloudflared
corgi db shell [name] # interactive DB shell inside running container (psql/redis-cli/mongosh/…)
corgi db shell [name] -e "SELECT 1" # one-shot query, exit; CI-friendly
corgi db snapshot [name] [svc] # fast physical Postgres snapshot (built indexes+matviews); postgres family only
corgi db restore [name] [svc] # restore pgdata from a snapshot → up on built data, zero recompute
corgi init # scaffold db_services/ + cloned repos
corgi create # interactive yml editor
corgi clean -i db # stop+remove db containers (also: services, corgi_services, all)
corgi pull # git pull in every service dir
corgi version # show installed version
corgi --describe # built-in: per-service JSON during parse; does NOT short-circuit. Use /corgi-describe for a rendered doc + Mermaid diagram
Full surface in references/commands.md.
Source: Andriiklymiuk/corgi — distributed by TomeVault.
