This really seems bananas. Why? What is the legitimate use case of the remote agent being able to run code on the local client's machine!?