bgreenwell/OctoAgent ? reverse-engineered prompt

Build me a Python tool that can take a GitHub repo and issue number, then use a small team of AI agents to figure out what the issue is, find the right file, propose a fix, review it a few times, and commit the change back to a branch in that repo.

I want it to run from the command line, with options to choose the repo owner, the issue, an optional target file if I already know it, how many review rounds to do, which model to use, and how much logging to show. It should read an OpenAI key and a GitHub token from environment variables, talk to the GitHub API, and keep the agent prompts in separate markdown files so they’re easy to tweak.

Please make it feel like an experimental but usable workflow, with clear console output, good error handling, and a clean project layout under src. If anything is unclear, look up the current OpenAI Agents SDK and GitHub API docs online if you need to.