Cost-aware delegation for Codex, with named specialist agents and an optional model guard. This standalone repository contains only the Codex-Orchestration plugin and its documentation.
An orchestrating agent can choose tasks dynamically while delegating through registered roles with fixed model and reasoning settings. The optional guard rejects covered spawn requests that violate the project policy. Enforcement starts off.
This public repository is MIT licensed and accepts community issue reports. It is a repository-hosted Codex catalog; an OpenAI public-directory listing and endorsement have not been granted.
| Goal | Guide |
|---|---|
| Understand the package and install it | Codex Orchestration quick start |
| Choose roles, toggle the guard, or add a custom agent | Configuration |
| Update, troubleshoot, or remove project assets | Operations and recovery |
| Understand what has actually been tested | Validation and compatibility |
| Contribute a change | Contributing |
| Report a security concern | Security |
| Report a bug or request an improvement | GitHub Issues |
First review the client and model prerequisites. These commands register the repository catalog and install its skill package:
codex plugin marketplace add https://github.com/Number531/Codex-Orchestration.git
codex plugin list --available --json --marketplace codex-orchestration
codex plugin add codex-orchestration@codex-orchestrationThe HTTPS catalog is publicly readable and requires no repository invitation. SSH is also supported for users with GitHub SSH authentication:
codex plugin marketplace add git@github.com:Number531/Codex-Orchestration.gitChoose one catalog source. If that catalog is already registered, inspect codex plugin marketplace list before adding it again. Plugin installation makes three skills available; it does not apply project configuration, grant hook trust, or enable enforcement. Continue with the project setup steps.
| Item | Value |
|---|---|
| GitHub repository | Number531/Codex-Orchestration |
| Codex marketplace name | codex-orchestration |
| Plugin ID | codex-orchestration |
| Package directory | plugins/codex-orchestration/ |
| Native catalog | .agents/plugins/marketplace.json |
Version 0.2.0 aligned the plugin, skills and project markers with this repository's community branding. Existing 0.1.0 installations need the manual migration steps. The generic role names, model pins and default-off behavior are unchanged. This repository contains no sibling plugins from the original multi-plugin repository.
For a reproducible install, add --ref <reviewed-ref> to the marketplace-add command above. Replace the placeholder with a reviewed commit SHA or a published tag from Releases; use one catalog source. No tagged release has been published yet. Review the upgrade and rollback procedure before changing an existing installation. Maintainers use the release procedure.
From the repository root, with Python 3.11+ and Git:
python3 -B -m unittest discover -s plugins/codex-orchestration/test -p 'test_*.py'
python3 -B plugins/codex-orchestration/test/check_catalog.py
python3 -B plugins/codex-orchestration/test/check_docs.pyThese checks use synthetic local fixtures and run in GitHub Actions. CI does not run paid models or native Codex sessions. Native engine checks, Desktop installation/trust, and live model access are separate validation surfaces; see the compatibility guide.
The repository and plugin use the MIT License. Individuals and organizations, including enterprises, may use, modify and redistribute the software commercially or noncommercially under those terms. Copies or substantial portions must retain the copyright and permission notices. The software is provided without warranty.
The installed plugin contains its own license copy, matching the repository license. GitHub write permissions remain controlled by the owner; the license governs reuse of copies.
Before a versioned release, verify license notices, source ownership and supported-client validation. Desktop UI and live-account qualification remain open. See maintainer release guidance. Publishing a GitHub release and submitting to OpenAI's public plugin directory are separate decisions.