ai-project-bootstrap
Docs / CLI reference / implement

implement

Writes a stack-tailored implementation plan, prompts, a validation checklist and skeleton files for one feature — not the whole project, and not the implementation either.

npx ai-project-bootstrap implement --list-features
npx ai-project-bootstrap implement authentication
npx ai-project-bootstrap implement authentication --dry-run

It never asks a question

It reads ai-project.config.json to see which technology answers the feature's category — authentication reads auth, payments reads payments — and writes content for exactly that combination.

Same command, different output

A project on Supabase Auth gets a plan opening with session persistence via AsyncStorage and the Row Level Security policies that gate access, plus an authClient.ts. The identical command on Clerk gets a plan opening with ClerkProvider and a secure token cache, no authClient — Clerk's own hooks are the client — and a useAuthedFetch.ts hook attaching a bearer token to your backend instead.

$ ai-project-bootstrap implement authentication

Implementing Authentication (Supabase Auth) in my-app…

Feature    Authentication — Supabase Auth
Plan       implementation/authentication/plan.md
Checklist  implementation/authentication/checklist.md
Prompts    1
    implementation/authentication/prompts/implement.md
Scaffold   4
    src/features/auth/authClient.ts
    src/features/auth/screens/SignInScreen.tsx
    src/features/auth/screens/SignUpScreen.tsx
    src/hooks/auth/useAuth.ts

Features and their providers

Feature
Reads category
Providers with real content
authentication
auth
Supabase Auth · Clerk · Auth0
payments
payments
RevenueCat · Stripe
push-notifications
notifications
Expo Push · OneSignal
No generic fallback

If your provider has no content for a feature, implement refuses with an error naming what is supported — rather than emitting generic filler with the provider's name substituted in.

Scaffold files are skeletons

A TODO and a pointer back to plan.md, not working implementations. You, or your assistant using the generated prompts, write the actual logic. Re-running is safe: a scaffold file you have hand-edited is left alone, tracked per feature in implementation/<feature>/.manifest.json.

Flags

Flag
Meaning
--list-features
Print every feature this version can implement
--dry-run
Preview without writing
--dir <path>
Run against a project elsewhere