In this post, I'll explore how AI skills can expose your agent to prompt injection, why seemingly safe skills still carry risks, and how to stay protected.