PromptArmor threat researchers uncovered a vulnerability in Anthropic's new Cowork that already was detected in the AI company's Claude Code developer tool, and which allows a threat actor to trick ...