Delegate a sub-task to OpenAI Codex CLI. Use when a task is better handled by Codex (e.g. complex file editing, multi-step coding tasks, automated code generation).
Scanned 5/27/2026
Install via CLI
openskills install Brainrot-Creations/claude-plugins---
name: delegate
description: Delegate a sub-task to OpenAI Codex CLI. Use when a task is better handled by Codex (e.g. complex file editing, multi-step coding tasks, automated code generation).
---
Use the `codex` MCP tools to delegate tasks to OpenAI Codex CLI.
## Prerequisites
Codex must be installed and signed in:
```bash
npm i -g @openai/codex
# sign in via the Codex app — no API key needed
```
## Synchronous delegation (need result now)
Use `codex_run` when you need the output before continuing:
```
codex_run(
prompt="<full task description with all context Codex needs>",
cwd="/path/to/project" # optional, defaults to current directory
)
```
Returns the final Codex message as a string.
## Background delegation (parallel work)
Use `codex_task_submit` to fire off tasks and continue working:
1. Submit one or more tasks:
```
codex_task_submit(name="Refactor auth module", prompt="...", cwd="...")
codex_task_submit(name="Write tests for API", prompt="...", cwd="...")
```
2. Check status:
```
codex_task_status(task_id="<id>")
```
3. Fetch result when completed:
```
codex_task_result(task_id="<id>")
```
4. See all tasks:
```
codex_tasks_list()
```
## Environment variables
| Variable | Purpose |
|----------|---------|
| `CODEX_PATH` | Override path to the `codex` binary |
| `CODEX_SANDBOX` | `read-only`, `workspace-write`, or `danger-full-access` — if unset, uses `--dangerously-bypass-approvals-and-sandbox` for headless operation |
## Writing good prompts for Codex
- Include full context — Codex starts fresh with no knowledge of prior work
- Specify the working directory if the task involves files
- Be explicit about what "done" looks like (files to edit, tests to pass, output expected)
- If the task spans multiple files, list the key ones
No comments yet. Be the first to comment!
Ultra-compressed communication mode. Cuts token usage ~75% by speaking like caveman while keeping full technical accuracy. Supports intensity levels: lite, full (default), ultra, wenyan-lite, wenyan-full, wenyan-ultra. Use when user says "caveman mode", "talk like caveman", "use caveman", "less tokens", "be brief", or invokes /caveman. Also auto-triggers when token efficiency is requested.