A file-based bridge that enables OpenClaw to dispatch, monitor, and manage complex coding tasks via the local Codex CLI.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install codex-bridge
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install codex-bridge 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 Bridge?
Codex Bridge is a sophisticated integration for Openclaw Skills designed to facilitate seamless delegation of heavy-duty coding tasks to a local OpenAI Codex CLI. It acts as an asynchronous execution layer, allowing a primary AI agent to offload script generation, code refactoring, and multi-step development work while remaining responsive for other user requests. By utilizing a robust file-based bridge, it ensures that long-running processes are tracked, managed, and resumed effectively through dedicated background execution paths.
Codex Bridge Use Cases
- Building complex automation scripts in Python, Bash, or JavaScript using Openclaw Skills.
- Implementing large-scale architectural refactors across an existing codebase.
- Offloading multi-file coding operations to background processes for better performance.
- Managing interactive coding tasks that require human clarification mid-execution.
- Maintaining continuous agent availability during intensive code generation cycles.
How Codex Bridge Works
- The user or agent dispatches a specific task using the dispatch script, defining the project directory and the prompt.
- The bridge initiates a background execution session via the local Codex CLI, isolating the process from the main agent loop.
- Openclaw Skills poll the task status periodically to capture real-time output, logs, or transition states.
- If Codex identifies an ambiguity, it pauses and writes a clarifying question to the bridge.
- Upon receiving an answer from the user via the bridge, the session resumes exactly where it left off.
- Once the task reaches a complete status, final code results and summaries are retrieved and presented to the user.
Codex Bridge Setup
To utilize this skill, ensure that python3 and the codex CLI are installed on your machine. You must also configure the execution permissions for the bridge scripts located within your Openclaw Skills directory.
# Ensure the scripts are executable
chmod +x ~/.openclaw/skills/codex-bridge/codex-bridge-dispatch.sh
chmod +x ~/.openclaw/skills/codex-bridge/codex-bridge-status.sh
chmod +x ~/.openclaw/skills/codex-bridge/codex-bridge-answer.sh
Codex Bridge Data Schema & Taxonomy
The bridge organizes task-related data within a structured local directory, ensuring Openclaw Skills can maintain state across asynchronous sessions.
| Component | Path / Detail |
|---|---|
| Task Root | ~/.codex-bridge/tasks/<task-id>/ |
| Status File | Contains the current lifecycle state (e.g., waiting_for_answer, running, complete). |
| Task Output | A continuous stream of CLI output from the Codex process. |
| Questions | Temporary storage for clarifying questions generated during the run. |
| Metadata | JSON records of task duration, working directory, and prompt parameters. |
Codex Bridge Advanced Features
- Background execution support enables non-blocking operations within Openclaw Skills.
- Interactive text-marker protocol for handling clarifying questions during autonomous runs.
- Automated 10-minute timeout for unattended tasks, ensuring sessions resume with sensible fallbacks.
- Detailed log and result inspection for debugging failed or errored coding executions.
- Full support for multi-agent workflows by isolating task states via unique task identifiers.
SKILL.md
Loading
Related Openclaw Skills
Pomodoro Timer
A terminal-based focus timer implementing the Pomodoro technique to enhance productivity through structured work and break intervals.
YouTube Summarizer Skill
Automatically fetch YouTube transcripts, generate structured summaries, and deliver them to messaging platforms.
Adversarial Prompting
A structured seven-phase methodology for solving complex problems through rigorous critique and iterative refinement within Openclaw Skills.
Visual Prompt Engine
A data-driven engine that creates high-quality, non-repetitive AI image prompts based on real-world design trends and structured visual vocabularies.
Torah Scholar
A comprehensive research tool for accessing the Sefaria library of Jewish texts directly within your AI agent workflow.
Canva API Automation Skill
A powerful integration for programmatically creating, managing, and exporting Canva designs via the Connect API.
Featured*
