Check

Check — Is This Project Ready to Run?

One pass tells you whether a repo is actually set up to run, and exactly what's missing if not. The deliverable is a ✓/✗ checklist across localhost / staging / prod, built by discovering the project's own expected setup — never by reciting generic advice.

North star: no one should discover a missing .env key or an unwired environment by hitting a runtime error — this surfaces it up front, with the fix.

Step 0 — Discover what SHOULD exist (never hardcode)

1. Env contract — .env.example (or env.example): the list of vars the app expects. This is the source of truth for "what must be set."
2. Environments doc — the repo's README/CLAUDE.md should describe how localhost, staging, and prod are each configured (compose file, secret manager, deploy target). Missing this is itself a ✗.
3. Run path — make help / grep -E '^[a-z-]+:.*##' Makefile, docker-compose*.yml, or package.json scripts: the real targets to set up and run locally.
4. Cloud auth — does the local stack need an SSO/CLI login before it can start (AWS/GCP/Azure profile)?

Step 1 — Localhost readiness

Report ✓/✗ for each:

• .env exists and is gitignored. ✗ → run the project's env-bootstrap target, or copy .env.example to .env.
• Every .env.example key is present in .env and non-empty. Diff the key sets — don't eyeball:

  comm -23 <(grep -oE '^[A-Z][A-Z0-9_]+' .env.example | sort -u) \
           <(grep -oE '^[A-Z][A-Z0-9_]+' .env 2>/dev/null | sort -u)
  

  Any line printed is a missing key (✗).
• No placeholder values left where a real one is needed (changeme, xxx, TODO).
• Local stack present — the run target exists and its deps are installed (Docker running? node_modules / venv present?).
• Cloud auth active if the local stack calls real cloud services.

Step 2 — Staging + prod wiring

For each non-local environment, confirm it's actually wired (don't assume parity with localhost):

• Every var the app reads has a value in that environment's config source — not just localhost. A var set locally but unset in staging/prod is a latent outage.
• Env-specific values (URLs, account IDs, buckets) are configured per environment, not hardcoded.
• Secrets exist in the secret store in the format the app expects, with a non-empty placeholder so an unset secret can't crash boot.
• The deploy/promotion path is documented and the running service references the current config.

This is a static check of the contract — it reads config sources, it does not deploy.

Step 3 — Report

Emit a compact checklist grouped by environment, ✓/✗ each, and for every ✗ the one command or edit that fixes it. End with a one-line verdict: ready to run locally? What blocks it? If there's no environments doc, recommend adding one — it's what makes this check repeatable next time.

Anti-patterns

• Assuming staging/prod parity from a green localhost — the unset-in-prod var is the classic outage.
• Eyeballing .env vs .env.example instead of diffing the key sets.
• Treating an undocumented environment setup as "nothing to check" — it means the contract is unwritten.
• Printing secret values in the report. Report key names and presence, never the value.