ClawdStrike

This skill is purpose-aligned and includes sensible anti-exfiltration rules, but it still requires immediate execution of a local collection script whose contents are not shown. That makes it moderately risky as an AI skill: more suspicious for execution trust than for malicious intent.