Yup, full access to DOM! Still needs a lot of optimizations, but the trick is that the agent reads the DOM as file, so it can grep parts of it naturally.