lockpick

SUSPICIOUS. The skill is largely coherent for a hosted website-security product and includes reasonable ownership and prompt-injection guardrails, but it gives an AI agent security-scanning capability and depends on an external `lockpicks` CLI whose official provenance was not verified from the provided evidence. I see no clear credential theft or malicious exfiltration, but the combination of external CLI trust gap plus agent-enabled security operations makes this a high security-risk skill rather than a benign documentation helper.