humanize

This skill is operationally consistent (parses input, uses an API key, calls a remote endpoint, returns results) but is suspicious because its explicit purpose is to help users bypass AI/plagiarism detectors. The primary risks are privacy/data-exfiltration (user text sent to a third party with unknown retention/training policies) and enabling unethical misuse (academic dishonesty, evasion of detection). There is no sign of traditional malware, obfuscation, or system-level backdoors, but the skill should be treated as suspicious and its use restricted/monitored. Recommend adding explicit warnings about acceptable use, privacy/retention details, and an option for local-only processing; consider rejecting or tightly gating 'bypass' functionality.