Security researchers from Zenity Labs have found that the open source AI agent OpenClaw can be completely hijacked through manipulated documents. All it takes is a hidden command in an ...