vercel-labs/agent-browser ? reverse-engineered prompt

Hey, can you build me a tool that lets an AI control a web browser? I want a simple command-line interface that can act as the 'hands and eyes' for an agent trying to complete a task on the web.

It should be able to do all the basic browser stuff: open a URL, get the content of the page, click buttons, fill in forms, and take screenshots. The core of this CLI needs to be really fast and efficient, so maybe build it in Rust.

The most important feature is making it easy for an AI to use. It should be able to get a structured view of the page, like an accessibility tree, so the agent can interact with elements by referencing simple IDs. Also, it would be amazing if it had its own built-in chat mode, so a user could just type log in to example.com with my credentials and the tool would figure out the steps. Make it easy to install via npm or Homebrew.