openmanus-ai/OpenManus ? reverse-engineered prompt

Build me a simple Python app called OpenManus that acts like a general purpose AI agent from the command line. I want to run it with one command, then type in an idea in the terminal and have it figure out the steps and help carry them out using an LLM API I configure in a local config file. It should be easy to set up with either a normal Python environment or uv, and it should come with a sample config I can copy and edit for my API key, model, and other settings.

Please make the project feel lightweight and practical, with a clean terminal flow, a main entry point, and a second more experimental flow if needed. If there are any missing details from the docs, look up the current best way to wire this kind of agent together. I’d also like the repo to be organized so it’s easy to extend later and contribute to.