A rigorous engineering specification for AI agents based on OpenAI Harness Engineering principles to ensure architectural integrity and task stability.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install codex-harness-engineering
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-harness-engineering 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 Harness Engineering?
Codex Harness Engineering is a specialized framework designed to optimize the performance of AI coding agents by focusing on the environment rather than just the model. By treating the repository as a system of record and implementing strict architectural constraints, this approach allows agents to handle complex, multi-file tasks with high reliability. Within the ecosystem of Openclaw Skills, it provides a standardized way to manage agent toolsets, documentation, and observability.
This skill emphasizes that slow progress in AI development is usually a result of insufficient environmental structure. By providing the necessary tools, abstractions, and versioned documentation, developers can create a closed-loop system where agents perform self-reviews and maintain the long-term health of the codebase without manual intervention.
Codex Harness Engineering Use Cases
- Orchestrating multi-step development tasks that span across multiple modules and files.
- Implementing automated PR review cycles where agents self-correct based on feedback.
- Driving development through observability using UI snapshots, system logs, and performance metrics.
- Preventing architectural drift in long-term projects through automated governance and technical debt tracking.
How Codex Harness Engineering Works
- Decompose large objectives into structured building blocks: design, implementation, review, and testing phases.
- Identify environmental deficiencies—such as missing tools or documentation—whenever an agent encounters a block.
- Feed discovered requirements back into the repository to build a cumulative, reusable skill set.
- Execute tasks via a strict layered architecture (Types > Config > Repo > Service > Runtime > UI).
- Use automated linters and CI checks to enforce architectural boundaries and document freshness.
- Implement periodic background tasks to scan for pattern drift and initiate refactoring PRs automatically.
Codex Harness Engineering Setup
To integrate Codex Harness Engineering into your Openclaw Skills workflow, establish the required directory structure and documentation maps:
# Create the standardized documentation hierarchy
mkdir -p docs/design-docs docs/exec-plans/active docs/exec-plans/completed docs/references
# Initialize the core agent map
touch AGENTS.md
# Configure local linting rules to enforce architectural layers
# Ensure the environment has access to gh (GitHub CLI) and git worktree
Codex Harness Engineering Data Schema & Taxonomy
The skill organizes repository data into a strictly versioned hierarchy to serve as the agent's source of truth:
| Component | Description |
|---|---|
| AGENTS.md | A concise map and directory for agent navigation; not an encyclopedia. |
| docs/exec-plans/ | Structured logs of active and completed execution strategies. |
| docs/tech-debt-tracker.md | A versioned document tracking architectural inconsistencies and drift. |
| docs/design-docs/ | Formal specifications that guide the agent's implementation logic. |
| Providers | Dependency injection points for cross-cutting concerns like logging and telemetry. |
Codex Harness Engineering Advanced Features
- Support for git worktree instances to enable parallel task execution in isolated environments.
- Advanced observability using LogQL and PromQL for SLO-based task execution and performance monitoring.
- DevTools integration providing DOM snapshots, automated screenshots, and navigation tracing.
- Automated throughput optimization that reduces blocking merge gates to shorten the PR lifecycle.
- Continuous pattern governance that uses automated quality scoring to trigger directional refactoring.
SKILL.md
Loading
Related Openclaw Skills
Notion
A comprehensive skill for programmatically managing Notion pages, databases, and block content via the latest API version.
Model Usage for CodexBar
Efficiently analyze and summarize per-model usage costs for Codex and Claude directly from your local CodexBar CLI data.
Team Weekly Report Summary
An AI-powered skill that transforms weekly report images into structured, comparative team summaries using OCR and intelligent classification.
AI Agent Cost Optimization Cookbook
A comprehensive technical guide and Python toolkit for measuring, managing, and drastically reducing the operational costs of AI agent fleets.
Heart Mender (Xinling Bushou)
Heart Mender is a pluggable AI persona module that injects customizable emotional support and flattery capabilities into any agent.
Apple Reminders CLI (remindctl)
A robust command-line interface for managing macOS Reminders, offering full CRUD capabilities and JSON support for advanced automation.
Obsidian
A professional integration for managing Obsidian vaults and automating Markdown note workflows via obsidian-cli.
Featured*
