init-workspace

init-workspace

Initialize any project for AI-assisted development with a single command.

The Problem

You're using multiple AI coding tools — Claude Code, OpenAI Codex CLI, Google Gemini CLI. Each has its own config file. You end up with:

• Instructions scattered across CLAUDE.md, AGENTS.md, GEMINI.md
• Each tool learning things the others don't know
• Inconsistent behavior when switching between tools
• Missing .gitignore entries that leak secrets or bloat repos
• No structure for custom skills, agents, or rules

The Solution

Run /init-workspace once. Get a project structure where:

• One file to rule them all: Write instructions in AGENTS.md only. Claude Code, Codex CLI, and Gemini CLI all read from it automatically.
• Shared memory: When any tool learns something new, it updates AGENTS.md. All tools stay in sync.
• Ready for customization: Directories for skills, agents, and rules are already in place.
• Safe defaults: Comprehensive .gitignore protects secrets and keeps repos clean.

Installation

Option 1: Plugin Marketplace (Recommended)

In Claude Code, run:

/plugin marketplace add xiaolai/init-workspace
/plugin install init-workspace@xiaolai-init-workspace

Then use /init-workspace:init-workspace to run.

Option 2: Personal Skill (Simpler)

mkdir -p ~/.claude/skills/init-workspace
curl -o ~/.claude/skills/init-workspace/SKILL.md \
  https://raw.githubusercontent.com/xiaolai/init-workspace/main/skills/init-workspace/SKILL.md

Or clone the entire repo:

git clone https://github.com/xiaolai/init-workspace /tmp/iw && \
  mkdir -p ~/.claude/skills/init-workspace && \
  cp /tmp/iw/skills/init-workspace/SKILL.md ~/.claude/skills/init-workspace/ && \
  rm -rf /tmp/iw

Then use /init-workspace to run.

Usage

In any project directory, run:

/init-workspace                         # if installed as personal skill
/init-workspace:init-workspace          # if installed as plugin

Claude will ask two questions:

1. Project description — A single line describing your project
2. Public or Private — Whether to share AI config files via git:
   • Public (recommended): Team shares the same AI context
   • Private: AI files are gitignored, each dev has their own

Then it creates:

project/
├── .claude/
│   ├── settings.json        # Team-shared settings
│   ├── agents/              # Custom subagents
│   ├── skills/              # Claude Code skills
│   └── rules/               # Modular rules
├── .codex/
│   ├── skills/              # Codex CLI skills
│   └── prompts/             # Codex CLI custom slash commands
├── .gemini/
│   ├── skills/              # Gemini CLI skills
│   └── commands/            # Gemini CLI custom slash commands (TOML)
├── .mcp.json                # MCP server configuration
├── .gitignore               # Comprehensive gitignore
├── AGENTS.md                # Shared instructions (all tools read this)
├── CLAUDE.md                # Claude Code (imports AGENTS.md)
└── GEMINI.md                # Gemini CLI (imports AGENTS.md)

What Gets Created

File/Directory	Purpose
AGENTS.md	Your single source of truth for all AI tools
CLAUDE.md	Imports @AGENTS.md (Claude Code reads this)
GEMINI.md	Imports @AGENTS.md (Gemini CLI reads this)
.claude/settings.json	Team-shared permissions
.claude/agents/	Custom subagents
.claude/skills/	Claude Code skills (slash commands)
.claude/rules/	Modular rules auto-loaded into context
.codex/skills/	Codex CLI skills
.codex/prompts/	Codex CLI custom slash commands
.gemini/skills/	Gemini CLI skills
.gemini/commands/	Gemini CLI custom slash commands (TOML)
.mcp.json	MCP server configuration
.gitignore	Comprehensive ignore rules

Customization

After initialization:

1. Edit AGENTS.md to add project-specific instructions
2. Add skills to .claude/skills/, .codex/skills/, or .gemini/skills/
3. Add subagents to .claude/agents/, rules to .claude/rules/
4. Configure MCP servers in .mcp.json

License

MIT

Sources

• Claude Code Skills Documentation
• Claude Code Plugins Documentation
• Codex CLI Skills Documentation
• Gemini CLI Skills Documentation
• AGENTS.md Open Format