Harness for Codex plugin
A Codex plugin that builds a team of subagents and skills for a project: an orchestrator splits tasks and checks the result.
- 68 ★ on the repository
- Last commit
- Apache-2.0
Extended access
Why this level
- The skill creates agent and skill files in the project, and the orchestrator runs subagents in parallel.
Install
- Works in
Manual install
codex plugin marketplace add https://github.com/revfactory/codex-harness.gitThe second README command: `codex plugin add codex-harness@codex-harness`.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The skill creates agent and skill files in the project, and the orchestrator runs subagents in parallel.
This is third-party code. Review the repository files before installing.
What it does
Harness for Codex is a harness factory for a single project. The `harness` skill inspects the project, designs a small team and creates native Codex agents and reusable skills. A parent orchestrator runs independent tasks in parallel, integrates results and owns final verification. The repository holds five agents, `harness-architect`, `harness-explorer`, `harness-worker`, `harness-reviewer` and `harness-qa`, plus six team architecture patterns. The project is an independent migration of `revfactory/harness` from Claude Code to Codex. Calling `$harness` in a session opened on a project builds or updates a durable harness in that project by default, even when the rest of the request is ordinary, such as an article. An explicit request to create nothing takes precedence.
Who it is for. Codex users who run a large project and want work split between specialised subagents.
Good fit when
- A new project needs a team of agents whose roles are designed from its structure.
- Independent tasks, such as reviewing several services, are better run in parallel.
- A service migration goes step by step and needs a separate agent to check contracts.
- A repeated workflow should become a project skill.
Not a fit when
- You need one small answer or edit: a team harness is overkill.
- You work in Claude Code: the upstream `harness` project exists for it.
- The Codex session was started outside the target project and cannot be restarted.
Example request
$harness Write a short blog post about organizing a new project and save it to `article.md`.Limitations
Version 0.1.0, commands verified on Codex CLI 0.153.4. Native subagent spawn has no working-directory parameter, so start Codex in the target project: `codex -C /absolute/path`. If the session began elsewhere, files can be written into the target project but the agents need a fresh session. `$harness` creates project files by default even when the task is not about a harness. Installing into another project with the scripts needs Python 3.11 or newer.
How to disable. Remove `codex-harness` with the `codex plugin` command, then delete the created `.codex/agents` and `.agents/skills` folders from the project if you no longer need them.
FAQ
How do I install the plugin in Codex?
Run `codex plugin marketplace add https://github.com/revfactory/codex-harness.git`, then `codex plugin add codex-harness@codex-harness`. You can also install it from `/plugins` inside the CLI.
How do I call the skill?
Open the target project in a new Codex session and send a request starting with `$harness`. `$codex-harness:harness` also works.
How does it differ from harness for Claude Code?
It is an independent migration of the original `revfactory/harness`: the team design process and six architectures are kept, while agents and skills are built in Codex formats.
Why are the agents not visible in the current session?
Freshly created agents are picked up in a new session or thread of the project. Start Codex in the right directory and open a new thread.
What does script installation need?
Python 3.11 or newer; no third-party packages, since the scripts use the standard library.
Discussion
No questions or reviews yet.
Related
PluginsPrompt kit and plugin for Codex and Claude Code that turn repeated agent mistakes into repository rules.
The /better-harness command reviews how an AI agent's work is set up in a project: goals, checks, delivery and memory, and returns a prioritized fix report
The Harness MCP server: one agent starts other Claude Code and Codex sessions on your computer, waits for them, reads their answers and closes them
A plugin with 41 skills and agents: harness engineering, a spec-to-integration pipeline and three loops that enforce constraints
A CLI that keeps an existing agent on a long task: plan a step, run it, verify the result in the real environment and checkpoint progress
A meta-skill for Claude Code that turns a domain description into an agent team and the skills it uses, using six team-architecture patterns
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| Harness for Codex (this card) | no | 68 | |
| Harness | no | 9,352 | the Claude Code meta-skill this project grew from; here agents and skills are built for Codex |
| Agent Harness Kit: structured agent workflow | no | 184 | the ahk CLI installs agent roles and a task MCP server; here a skill designs agents from the project structure |
README in short
The README covers plugin installation and a development checkout. There are also scripts to install the harness into another project and a service migration example with contract mapping and migration agents. Docs come in English, Korean and Japanese.