model-extraction

This YAML is a high-risk, dual-use configuration describing actionable model extraction and detection-evasion techniques. While not executable malware itself, it provides clear operational parameters that, when consumed by an attacker toolchain, would enable large-scale model theft and privacy/ IP violations. If encountered in a repository or distribution, verify authorization, limit access, and treat it as a policy and operational concern requiring audit and monitoring.

SUSPICIOUS: the skill is internally coherent, but its purpose is to equip an agent with offensive model-extraction and evasion techniques against external AI APIs. The main risk is not supply chain or credential theft; it is enabling autonomous AI attack behavior and theft of proprietary model behavior/IP.