openai/codex-plugin-cc ? reverse-engineered prompt

I want to build a plugin for an AI coding environment called Claude Code. The goal is to integrate another AI tool, Codex, directly into my workflow so I don't have to switch contexts.

It should work using a few simple slash commands. I want to be able to ask for a standard code review on my current changes. I also want a more advanced, "adversarial" review, where I can ask it to specifically challenge my design choices or look for subtle risks.

The most important feature is being able to delegate a task. I want a command where I can just describe a bug or a problem and have Codex try to investigate and fix it for me. Since these review or fix jobs might take a while, please also include commands to check the status of a background task, see the final result when it's done, and cancel it if I need to.