smart-init

name	smart-init
description	Generate optimized CLAUDE.md by analyzing the codebase. Use when starting a new project or when /init is called. Detects stack, patterns, and creates production-ready project config.

Smart Init - Automatic CLAUDE.md Generator

When the user runs /init or asks to set up the project, analyze the codebase and generate a complete, optimized CLAUDE.md file automatically.

Analysis Steps

1. Detect Tech Stack

• Check package.json, requirements.txt, pom.xml, Cargo.toml
• Identify frameworks (React, FastAPI, Spring Boot, etc.)
• Find build tools (npm, poetry, maven, cargo)

2. Detect Project Structure

• Find main source directories (src/, app/, lib/)
• Identify test directories (tests/, __tests__/, test/)
• Locate config files

3. Detect Commands

• Extract scripts from package.json, Makefile, pyproject.toml
• Identify common patterns (dev server, tests, build, lint)

4. Detect Conventions

• Check for linter configs (.eslintrc, ruff.toml)
• Find type checker configs (tsconfig.json, mypy.ini)
• Identify existing patterns in code (naming, imports)

CLAUDE.md Generation Template

Generate this structure, filling ALL fields automatically:

# [Project Name from package.json/pyproject.toml]

## About
[1-2 line description from package.json description or README]

## Stack
**Languages**: [Detected: Python 3.13, TypeScript, etc.]
**Frameworks**: [Detected: FastAPI, React, Spring Boot, etc.]
**Database**: [Check for postgres, mysql, mongodb in deps]
**Tools**: [Detected: Docker (if Dockerfile exists), Git]

## Structure
\`\`\`
[Actual directory tree of main folders]
\`\`\`

## Commands
\`\`\`bash
# Development
[Extracted from package.json scripts or detected patterns]

# Testing
[Found test commands]

# Build/Deploy
[Found build commands]
\`\`\`

## Code Standards
[Based on detected linter/formatter configs]
- [Language]: [Conventions from existing code]
- [Framework]: [Patterns found in codebase]
- [Testing]: [Test patterns detected]

## Git Workflow
- Branch: `[feature|fix|docs]/description`
- Commits: [If commitlint detected: "Conventional Commits", else: "Descriptive messages"]

## Key Behaviors
1. Read only files explicitly mentioned or necessary
2. Make minimal, focused changes
3. Run tests after code changes
4. Explain approach for complex tasks

## Ignore
- Dependencies: [List actual dependency dirs found]
- Generated: [List actual build dirs found]

---
*Auto-generated by smart-init. Edit as needed.*

Smart Defaults

If something cannot be detected, use intelligent defaults:

For Python projects:

• Structure: /src, /tests, /docs
• Commands: pytest, ruff check, mypy
• Standards: Type hints required, async/await for I/O

For JavaScript/TypeScript:

• Structure: /src, /tests, /public
• Commands: npm run dev, npm test, npm build
• Standards: ESM imports, strict TypeScript

For Java:

• Structure: /src/main/java, /src/test/java
• Commands: ./mvnw test, ./mvnw spring-boot:run
• Standards: Constructor injection, no field injection

For any project:

• Git: Conventional commits
• Testing: Arrange-Act-Assert pattern
• Ignore: Standard dependency/build directories

Output Format

1. Analyze the codebase (use bash, read files)
2. Generate complete CLAUDE.md with NO placeholders
3. Save to .claude/CLAUDE.md
4. Show summary of detected configuration
5. Ask if user wants to customize anything

Example Output Message

✅ Generated .claude/CLAUDE.md

Detected:
- Stack: Python 3.13 + FastAPI
- Framework patterns: Async/await, Pydantic models
- Tests: pytest (12 test files found)
- Commands: 5 npm scripts extracted

The file is ready to use. Run 'view .claude/CLAUDE.md' to review it.

Want to customize anything?

Critical Rules

• NEVER leave [placeholder] or [TODO] in the generated file
• If you can't detect something, use smart defaults from above
• The generated file must be immediately usable
• Keep total length <50 lines
• Use actual values from the codebase, not examples