arthai-marketplace

/setup

Bootstrap any project from empty repo to deployed app. Modular — run full or target a module.

Synopsis

/setup [module]

When to use it

Starting a new project and want the full path: repo → deps → Docker → database → credentials → CI → deploy
A specific piece is broken or missing — run just that module (/setup credentials, /setup infra, …)
Checking how far along a project’s setup is — /setup status
Not for: teaching the toolkit about an existing codebase — that’s /calibrate; or restarting dev servers — that’s /restart

Quickstart

/setup

What you’ll see: the project’s stack is detected (from CLAUDE.md and package/config files), then modules run in order — skipping anything already complete — with a status check after each. Every module is idempotent: re-run safely to fix or verify. “Already complete” is decided per module by a concrete check — e.g. .git/ existing (repo), a running Docker container (infra), a populated migrations table (database), gh CLI + auth present (secrets sync) — not a guess; see /setup status for the same checks surfaced as a table.

Examples

/setup                  # full interactive setup — detect what's needed, skip what's done
/setup status           # readiness table across all 16 modules
/setup prerequisites    # system requirements + account creation
/setup repo             # git init + .gitignore + project scaffolding
/setup deps             # install dependencies (venv, npm, cargo, ...)
/setup infra            # start Docker services (DB, cache, queue)
/setup database         # run migrations + seed data
/setup credentials      # API keys + env files, with per-key validation
/setup profiles         # bot/service identity + social presence
/setup verify-local     # start the app + smoke tests
/setup claude-agents    # install the toolkit + populate CLAUDE.md
/setup arth             # register the project with the Arth dashboard
/setup ci               # generate a GitHub Actions workflow for the stack
/setup discord          # Discord server + bot + webhooks + ChatOps
/setup secrets-sync     # push secrets to GitHub + cloud provider
/setup deploy           # cloud deployment (Railway, Vercel, AWS, ...)
/setup verify-prod      # production smoke tests
/setup monitoring       # Sentry / error tracking setup

What it does

Discovers project context — reads CLAUDE.md and scans package.json, requirements.txt/pyproject.toml, go.mod, Cargo.toml, docker-compose.yml, env files, migration dirs, and deploy configs to adapt every module to your stack.
Prerequisites — checks required software (Python/Node/Go/Rust/Docker/git/gh/CLI tools) and detects required accounts from .env.example (Anthropic, OpenAI, Stripe, Stytch, …) with a status table.
Repo, deps, infra, database — git init + .gitignore + scaffolding per stack; language-specific installs (incl. ML model pre-downloads); Docker services with per-service health checks; migration tool detection (Alembic/Prisma/Django/Knex/Drizzle) + seed data.
Credentials (user-confirmation checkpoint) — creates the env file from the template, auto-generates secret keys, then guides you through each missing credential with get/cost/validate steps. You paste values yourself — it never stores or echoes credentials.
Local verify, toolkit, CI — starts the app, hits health endpoints, runs tests; installs the claude-agents toolkit and populates CLAUDE.md via /scan; generates a stack-matched GitHub Actions workflow with fake test env vars.
Discord, secrets sync, deploy — optional Discord server/bot/webhooks for CI + ChatOps; pushes real secrets to GitHub and the cloud provider — always confirms with you before pushing to remote platforms (a user-confirmation checkpoint); deploys to the detected platform (Railway/Vercel/Fly/AWS/GCP) with post-deploy migrations and webhook registration.
Prod verify + monitoring — production health checks, smoke tests, log review; optional Sentry setup.
Completion verification — spawns an independent verifier that reports PASS / GAPS FOUND / INCONCLUSIVE. Gaps don’t block; you decide whether to rerun.

Agents spawned

Agent	Model tier	Role
completion-verifier	sonnet	Independent check that setup actually completed

Output & artifacts

Project files: .gitignore, scaffolding, env file, docker-compose services running, migrations applied, .github/workflows/ci.yml, deploy configuration
CLAUDE.md populated and the claude-agents toolkit installed (module 8)
Secrets in GitHub Secrets + cloud provider env vars (after your confirmation)
/setup status — a 16-row readiness table (✅/❌/⚠️ per module)
A context tip on completion: everything is persisted to files — start a fresh session afterwards (and run /calibrate there)

Troubleshooting

Problem	Fix
`ValidationError` on app startup	Missing env vars — run `/setup credentials`
`ConnectionRefusedError` on :5432 / :6379	Postgres/Redis not running — `docker compose up -d`
`alembic upgrade` fails: tables already exist	`alembic stamp head`, then re-run the migration
`npm ERR! ERESOLVE`	Dependency conflict — `npm install --legacy-peer-deps`
CI fails on service health	Container not ready — add health-check options to the workflow
`install.sh` permission denied	`chmod +x ~/.claude-agents/install.sh`

/calibrate — deep-learn the project you just bootstrapped (run it next, in a fresh session)
/scan — the CLAUDE.md population step used by module 8
/deploy — day-to-day deploys after first-time setup
/restart — restart and validate local dev servers