obliteratus

SUSPICIOUS. The skill is internally coherent, but its stated goal is to remove LLM safety controls, which is a high-risk capability for an AI agent. The install path uses an unpinned editable install from a personal GitHub repo, and optional telemetry adds minor data-flow uncertainty. I do not see confirmed credential theft or hidden exfiltration, so this is better classified as high-risk vulnerable/suspicious rather than malware.