What is sol-governed-codex?

Sol-Governed Codex is a toolkit that plugs into OpenAI Codex to organize how multiple AI agents work together on coding tasks, while keeping an eye on cost. Instead of using the most expensive, most capable AI model for every single edit, it splits work across three roles and only calls in a stronger reviewer when the task actually needs one. The three roles are Terra, Luna, and Sol. Terra sits at the root and owns requirements, integration, and talking to the developer. Luna is a low-cost worker meant for mechanical, well-defined edits and tests. Sol is a bounded advisor brought in only for higher-stakes moments, such as reviewing a risky plan, architecture decisions, security, billing, or destructive operations, and it gives back a clear verdict of PASS, REVISE, or BLOCKED along with supporting evidence rather than a vague thumbs up. Setup requires Git, Bash, Python 3.11 or later, and an existing OpenAI Codex installation. It can be installed either through the Codex plugin marketplace with a couple of commands, or by cloning the repository directly and running an installer script that copies the skill files and agent profiles into your Codex home folder. The installer includes safety checks: it will not silently overwrite files you have modified unless you explicitly force it, and it refuses to replace an existing AGENTS.md file in your project without your review. Once installed, a typical workflow looks like starting Codex with one of two provided profiles, giving the root agent a clear goal and acceptance criteria, letting Luna and Terra handle the actual work at their appropriate skill levels, and calling on Sol only for the moments that matter, like checking a diff before something risky ships. There is a validation script to confirm the setup is working correctly without spending on an actual model call. The project is MIT licensed and includes automated validation checks visible in its GitHub Actions badge.