hunt-llm-ai

SUSPICIOUS. The skill is internally coherent as an offensive AI-security hunting guide, but its footprint is high-risk: it equips an AI agent to conduct prompt-injection exploitation, OOB exfiltration, SSRF to metadata services, and cross-tenant data theft validation against real systems. No strong malware evidence or hidden execution appears, but the combination of exploit guidance, known exfiltration endpoints, and autonomous external actions makes this a dangerous skill.