iBz-04/raya — reverse-engineered prompt

Reverse engineered prompt

GitHub

Build me a Windows desktop AI agent in Python that can take a plain English request and actually use my computer to do it. I want it to be able to open apps, click buttons, type text, run commands, and read the current UI state so it can finish simple tasks for me, like writing a short note and saving it to the desktop, or switching from dark mode to light mode.

Make it usable both from a small script and from the terminal with a main entry file. It should let me plug in an LLM with an API key from an env file, and return a clear final result after the task runs. If there is an option to use browser control or vision, keep that available.

Please keep the project installable as a package, add basic tests, and include a short README with setup and example prompts. This is for Windows, and it should be careful about unsafe actions and make it obvious that automation can make mistakes. Look up current docs online if you need to.

Want more depth? Deep Reverse