Lazy Claudecode

Make Claude Code work like it means it.

Built because Codex hit its limits and we needed something that doesn't stop working.
Inspired by LazyCodex by @code-yeongyu — rebuilt from scratch on Claude Code's native hook system.

🚀 Quick Start

# Requires Bun (https://bun.sh)
curl -fsSL https://bun.sh/install | bash

npm install -g lazy-claudecode
lazy-claudecode install

Verify everything is set up:

lazy-claudecode doctor

That's it. Next time you open Claude Code, the hooks are active.

git clone https://github.com/effortprogrammer/lazy-claudecode.git
cd lazy-claudecode
bun install && bun link
lazy-claudecode install

⚡ Ultrawork Mode

Add ulw anywhere in your prompt:

> implement the auth module ulw

Claude will:

1. Explore the codebase to understand what exists
2. Plan a decision-complete work breakdown
3. Execute with a test → code → verify loop
4. Review the result with parallel QA agents
5. Loop until everything actually works — no babysitting needed

🤝 Cross-Call: Claude Code ↔ Codex

If you have both CLIs installed, the installer sets up bidirectional delegation automatically.

Claude Code → Codex

Use the /lazycodex slash command inside Claude Code:

/lazycodex refactor the auth module to use dependency injection

Codex → Claude Code

Use the $lazycc skill inside Codex:

$lazycc review this PR for security issues

Note: Claude Code uses /command syntax, Codex uses $skill syntax — that's just how each tool works.

Each agent runs the other's CLI under the hood and returns the result as-is — no context lost, no copy-paste.

Requirements

If the target CLI is missing, the command shows installation instructions instead of failing silently.

🧬 Origins

Built by Hojin Yang. Inspired by LazyCodex (@code-yeongyu).

License

MIT