ai-jailbreak

SUSPICIOUS. The skill is internally consistent and contains no installs, credentials, or exfiltration paths, but its purpose is explicitly to help an agent bypass LLM safety controls. That offensive capability makes it high security risk despite low evidence of malware.