The agent had a hard time with browser use. All i asked it to do was try the chat input. It initially got only a shallow screenshot with the input below the fold. I instructed it to scroll down, then it could see the input, but it had a hard time using the input.
(But thank you for building this, I’m very excited about it)
