Automate long-running AI coding tasks using persistent tmux sessions and self-healing retry loops.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install coding-agent-loops
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 coding-agent-loops 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 Coding Agent Loops?
Coding Agent Loops provide a robust framework for managing autonomous AI developers like Codex and Claude Code. By utilizing Openclaw Skills, developers can wrap agentic workflows in persistent tmux sessions that survive system restarts and network interruptions. This pattern ensures that instead of a single, fragile session that might stall due to context bloat, the agent runs in focused iterations, picking up progress via git history and local files.
At its core, this skill implements the Ralph loop pattern, which prioritizes short, high-intent sessions over long-lived ones. This approach is essential for modern software engineering where Openclaw Skills are used to bridge the gap between AI generation and reliable code execution in local environments.
Coding Agent Loops Use Cases
- Executing multi-step feature developments based on PRD checklists using Openclaw Skills.
- Running autonomous agents for tasks prone to stalling or requiring high-context processing.
- Managing parallel AI agents across different repositories or tasks simultaneously.
- Ensuring coding tasks finish and notify the user via Openclaw Skills even if the local machine restarts or the terminal is closed.
How Coding Agent Loops Works
- Initialize a persistent tmux session using a stable socket to prevent session reaping by the operating system.
- Launch the AI agent (Codex or Claude) through the ralphy-cli wrapper, which manages the lifecycle of the task.
- The agent processes tasks based on a PRD or direct prompt, tracking progress via git and markdown checklists.
- If the agent fails or stalls, the Ralph loop pattern restarts the process with fresh context to overcome technical hurdles.
- Upon task completion, a mandatory hook triggers an Openclaw Skills system event to notify the user immediately.
- The session remains open for manual verification of git logs and diffs before the developer closes the loop.
Coding Agent Loops Setup
To get started with these Openclaw Skills, ensure you have tmux and the necessary CLI tools installed:
# Install the Ralph loop runner
npm install -g ralphy-cli
# Ensure tmux is installed and create a stable socket directory
mkdir -p ~/.tmux
# Authenticate your preferred agent
codex auth login
Integration with Openclaw Skills requires appending the completion hook to your tmux commands to capture exit codes and fire system events correctly.
Coding Agent Loops Data Schema & Taxonomy
The skill organizes its operational data and state using the following structure to ensure compatibility with Openclaw Skills:
| Component | Description |
|---|---|
| TMUX Socket | Located at ~/.tmux/sock to ensure persistence across system-level socket cleanup. |
| PRD Checklist | Markdown files using [ ] and [x] to track agent progress and completion state. |
| Session Logs | Accessible via tmux capture-pane or stored in ~/.codex/log/ for debugging. |
| Git History | Serves as the primary source of truth for the agent's progress and file state. |
Coding Agent Loops Advanced Features
- Multi-agent parallel execution using the
--parallelflag for rapid feature builds across multiple Openclaw Skills workflows. - Integration with Openclaw Skills system events for real-time desktop or mobile notifications upon task success or failure.
- Self-healing loops that automatically retry on API rate limits (429 errors) without manual intervention.
- Custom completion hooks to chain multiple AI workflows or automated post-processing scripts after the agent finishes.
SKILL.md
Loading
Related Openclaw Skills
Whoo CLI
A specialized CLI tool for fetching and interpreting personal health metrics from the WHOOP API.
Plane.so CLI
A zero-dependency Python CLI for managing Plane.so projects, issues, and team workflows directly from your terminal.
Zillow Scraper
A high-performance real estate data extraction tool for scraping Zillow listings, price history, and market analytics while bypassing Cloudflare protection.
YouTube Competitor Analysis
A powerful tool for analyzing YouTube competitor channels, trending content, and search keywords to reverse-engineer successful video strategies.
Coding Sessions
Orchestrate persistent AI coding agents in tmux sessions with automatic notifications and health monitoring.
ElevenLabs Phone Calls
A sophisticated tool for making outbound AI-driven phone calls using ElevenLabs Conversational AI agents and Twilio integration.
Featured*
