Codex Autopilot for Openclaw

An automated multi-project orchestration system for Codex CLI using tmux and launchd on macOS.

imwyvern
v0.5.0
Mar 4, 2026
0
0
0

Install & Download

1. ClawHub CLI

The fastest way to install a skill directly from the registry.

npx clawhub@latest install autopilot

2. Manual Installation

Copy the skill folder to one of these locations

Global
~/.openclaw/skills/
Workspace
<project>/skills/

Priority: Workspace > Local > Bundled

3. Prompt Installation

Copy this prompt to OpenClaw to install it automatically.

Help me install autopilot using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).

Prefer to download?

Get the raw skill files in a ZIP archive.

What is Codex Autopilot?

Codex Autopilot is a sophisticated automation layer designed for developers managing multiple concurrent AI-assisted coding sessions. By leveraging tmux for session multiplexing and launchd for scheduled execution, it provides a robust watchdog system that keeps AI agents productive without manual intervention. This Openclaw Skills entry highlights how the system maintains project momentum by auto-nudging idle sessions and handling permission prompts.

The framework acts as a bridge between the CLI-based AI agent and the developer's operational needs, providing features like automated code reviews, PRD verification, and state management. It is particularly valuable for complex development environments where keeping an AI agent focused across multiple project directories is a core requirement.

Codex Autopilot Use Cases

  • Managing multiple concurrent Codex CLI coding sessions across different repositories.
  • Automating background development tasks like code reviews or PRD verification.
  • Orchestrating parallel AI-assisted coding workflows without manual supervision.
  • Receiving real-time status updates via Discord or Telegram for long-running AI tasks.

How Codex Autopilot Works

  1. The watchdog loop initiates via launchd and scans configured project tmux windows.
  2. The system analyzes terminal output to identify if a session is active, idle, or waiting for user permissions.
  3. Based on pre-defined safety limits, the autopilot dispatches actions such as auto-nudging idle prompts or approving file permissions.
  4. Tasks are dequeued from a prioritized list and sent to the appropriate tmux pane using dedicated injection scripts.
  5. Status information is synced to a central JSON state file and notifications are pushed to external communication channels.

Codex Autopilot Setup

First, clone the repository and initialize your configuration file:

git clone https://github.com/imwyvern/AIWorkFlowSkill.git ~/.autopilot
cd ~/.autopilot
cp config.yaml.example config.yaml

Install the required system dependencies via Homebrew:

brew install tmux yq jq

Register the launchd agent to start the background watchdog:

./install.sh

Codex Autopilot Data Schema & Taxonomy

This skill utilizes a structured file hierarchy to manage state and project configurations as part of the Openclaw Skills ecosystem:

Component Format Description
config.yaml YAML Global settings for timing thresholds, safety limits, and project directories.
projects/ YAML Per-project task definitions and priority levels.
state/ JSON Runtime state tracking for active sessions and loop detection.
logs/ Text Rotating logs for debugging watchdog actions and task execution.
task-queue/ Filesystem Persistent storage for pending and processed task IDs.

Codex Autopilot Advanced Features

  • Multi-project scheduler supporting both round-robin and priority-based task dispatching.
  • Integrated loop detection and daily token usage tracking to prevent runaway AI costs.
  • Claude fallback handler to maintain workflow continuity if the primary Codex service is unavailable.
  • Automated PRD verification engine that audits the codebase against requirement documents.
  • Discord and Telegram integration for remote monitoring and mobile status alerts.

SKILL.md


Loading

METADATA

Github Stars: 0
forks: 0

Featured*