| name | clone-implementation |
| description | Scaffold project and implement MVP from spec -- boilerplate, core features, AI integration |
Clone Implementation
When to Use
- Technical specification ready (output of clone-spec-writer)
- Time to write actual code
- Goal: working MVP, not perfect product
Process
- Scaffold project -- generate boilerplate:
- Initialize repo with chosen framework (Next.js, FastAPI, etc.)
- Configure linters, formatters, pre-commit hooks
- Set up environment variables template (.env.example)
- Create folder structure matching spec's component map
- Implement data layer -- database first:
- Define models/schemas (SQLAlchemy, Prisma, etc.)
- Write and run migrations
- Seed development data
- Add CRUD repository layer
- Implement API -- endpoints from spec:
- Auth flow (JWT/OAuth)
- Core business endpoints
- Input validation (Pydantic, Zod)
- Error handling middleware
- Implement UI -- screens from spec:
- Layout components (header, sidebar, main)
- Page components per route
- Forms with validation
- State management (zustand, Redux, etc.)
- Integrate AI components -- the GenAI core:
- LLM provider client (Gemini, OpenAI, Anthropic)
- Prompt templates with variable injection
- Streaming response handling
- Fallback chain (primary model --> fallback model)
- Wire together -- connect all layers:
- Frontend calls API, API calls AI, results render in UI
- End-to-end smoke test of critical path
Output
- Working MVP codebase in git repository
- README with setup instructions
- .env.example with all required variables
- Development seed data
- At least one end-to-end flow working (e.g. "create presentation from prompt")
Example
SlideDeck AI MVP: Next.js frontend scaffolded with 8 pages, FastAPI backend with 22 endpoints, PostgreSQL with 6 tables, Gemini integration for slide content generation. Critical path works: user enters topic --> AI generates outline --> slides render in editor --> export to PDF.