ai-model-extraction

SUSPICIOUS. The skill is coherent and contains no hidden installs or exfiltration paths, but its purpose is to enable offensive extraction of models, training data, and system prompts from AI endpoints. This makes it a high-impact security technique skill, though not malware.