init-workspace

# Initialize Multi-Agent Workspace

Create shared configuration files for Claude Code, Codex CLI, and Gemini CLI, plus the complete Claude Code project structure.

## Instructions

1. Ask the user two questions:
   - "Single line description of your project:"
   - "Keep AI config files private or public?"
     - **Private**: AI files (AGENTS.md, CLAUDE.md, GEMINI.md, .claude/, .mcp.json, dev-docs/) are gitignored. Use when you don't want to share AI instructions with collaborators.
     - **Public** (recommended): AI files are committed to the repo. Use when you want the team to share the same AI context.

2. Create the following directory structure and files:

### Directory Structure

```
project/
├── .claude/
│   ├── settings.json        # Team-shared settings
│   ├── agents/              # Custom subagents
│   │   └── .gitkeep
│   ├── skills/              # Custom skills
│   │   └── doc/
│   │       └── SKILL.md     # /doc command
│   └── rules/               # Modular rules
│       └── .gitkeep
├── dev-docs/                # Generated documentation
│   └── .gitkeep
├── .mcp.json                # MCP server configuration
├── .gitignore               # Comprehensive gitignore
├── AGENTS.md                # Shared instructions (Codex CLI native)
├── CLAUDE.md                # Claude Code (imports AGENTS.md)
└── GEMINI.md                # Gemini CLI (imports AGENTS.md)
```

### File Contents

#### AGENTS.md
```markdown
# Project Instructions

> $USER_RESPONSE

## Guidelines

<!-- Add your project-specific instructions here -->

## Shared Memory

**Always write new instructions, rules, and memory to `AGENTS.md` only.**

Never modify `CLAUDE.md` or `GEMINI.md` directly - they only import `AGENTS.md`.
This ensures Claude Code, Codex CLI, and Gemini CLI share the same context consistently.

## Project Structure

- `.claude/agents/` - Custom subagents for specialized tasks
- `.claude/skills/` - Slash commands (e.g., `/doc`, `/test`)
- `.claude/rules/` - Modular rules aut