It’s barely been out for a month and already security researchers have discovered a prompt injection vulnerability in Google’s Gemini command line interface (CLI) AI agent that could be exploited to ...