apply-guidelines

SUSPICIOUS leaning medium-risk, not malicious. The skill’s purpose matches its capabilities, and data flows stay within GitHub, local files, and project-defined checks. The main concern is trust: it fetches unpinned, mutable instructions from a personal GitHub repo on every run, then may apply those instructions to local code and execute project commands. That is a real supply-chain risk for an AI skill, but there is no clear credential harvesting, exfiltration, covert behavior, or malicious pre-execution payload.