A skill is someone else's code and prompts running inside your agent. Attackers have figured that out, and malicious skills are already turning up in public skill directories.
New research from Starseer Labs takes both sides. First we build malicious agent skills from the
