browser-use/browser-harness ? reverse-engineered prompt

I want to build an AI agent that can use my computer's web browser to do things for me. I found this project, Browser Harness, on GitHub that looks like what I need.

The main idea is that the agent should be "self-healing." This means if it gets stuck on a task because it doesn't know how to do something, like upload a file or handle a weird popup, it should be able to write the code it needs to learn that new skill and then keep going. The goal is to fully automate any browser task.

I'd like to be able to give it high-level commands like "Go to LinkedIn and find me three marketing jobs in Austin" or "Log in to my GitHub and star this repository for me." It should connect directly to my real Chrome browser. The repo has an install.md file that explains the setup, so please read that first.