Scaffold Project (CLI)¶

See how to bootstrap a new project from a PRD or README. SkillMeat analyzes your documentation, selects matching artifacts, and generates a ready-to-use .claude/ directory.

About This Demo

Duration: ~60 seconds
Audience: Developers starting new projects
What you'll see: PRD analysis, scaffolding, generated project structure, and collection integrity

Scaffold Project CLI Demo

What You'll See¶

The Source: Your PRD¶

Start with a plain Markdown PRD describing your project goals and tech stack.

head -30 project-prd.md

What's happening: - SkillMeat reads plain Markdown — no special schema needed - The PRD includes title, goals, and explicit tech stack (FastAPI, SQLite, pytest) - This is all the scaffolder needs to detect patterns and select artifacts

Running the Scaffolder¶

One command analyzes your PRD and generates the project structure.

skillmeat scaffold \
  --from-context project-prd.md \
  --project ./my-project \
  --auto-confirm \
  --collection my-collection

What's happening: - --from-context accepts a file path, directory, or free-text description - --project sets the target; scaffold writes into .claude/ there - The analyzer detects Python/FastAPI stack from the PRD content - --auto-confirm makes it non-interactive — safe for CI and reproducibility

Viewing Generated Structure¶

Check what was created in the project's .claude/ directory.

ls -la ./my-project/.claude/

What's happening: - .claude/CLAUDE.md is pre-populated with the detected architecture - Skills, commands, and agents are linked in place — already wired for your stack - Everything is ready to use immediately - No manual setup or configuration needed

Collection Remains Unchanged¶

The scaffolder reads from your collection but doesn't modify it.

skillmeat list --collection my-collection

What's happening: - Your collection inventory is exactly the same as before scaffold - Scaffold targets the project, not the collection - The same collection can scaffold any number of projects - No side effects — only the project directory is affected

What Scaffold Detects¶

The analyzer looks for language, framework, and tooling clues:

Pattern	Detected Framework
`FastAPI`, `async def`, `Pydantic`	Python/FastAPI backend
`React`, `useState`, `useEffect`	React/TypeScript frontend
`pytest`, `test_`	Python testing framework
`jest`, `describe()`	JavaScript/TypeScript testing
`SQLite`, `migrations`	Data layer patterns

Input Formats¶

The --from-context flag accepts multiple formats:

# From a PRD file
skillmeat scaffold --from-context ./PRD.md --project ./my-project

# From a directory (reads all .md files)
skillmeat scaffold --from-context ./docs --project ./my-project

# From free-text description
skillmeat scaffold --from-context "FastAPI + SQLite backend with pytest" --project ./my-project

# From a remote repo (planned)
skillmeat scaffold --from-context https://github.com/user/repo --project ./my-project

Key Takeaways¶

PRD is the source: Plain Markdown — no special schema needed
Automatic detection: Tech stack is inferred from content
Non-destructive: Collection is unchanged; only the project is created
Immediate setup: Generated .claude/ is ready to use right away
Scriptable: --auto-confirm flag makes it safe for automation

Try It Yourself¶

# Create a simple PRD
cat > my-prd.md << 'EOF'
# My Task Tracker API

Build a FastAPI backend for task management.

## Stack
- FastAPI for HTTP API
- SQLite for data storage
- pytest for testing

## Features
- CRUD endpoints for tasks
- User authentication
- Rate limiting
EOF

# Scaffold the project
skillmeat scaffold --from-context ./my-prd.md --project ./my-project --auto-confirm

# Check what was generated
ls -la ./my-project/.claude/

# Start building!
cd ./my-project
cat CLAUDE.md

Common Flags¶

Flag	Purpose
`--from-context <path/text>`	Source: PRD file, directory, or free text
`--project <path>`	Target project directory
`--collection <name>`	Source collection (default: active)
`--auto-confirm`	Skip confirmations (safe for scripts)
`--no-llm-analyzer`	Use pattern matching only (no LLM)