verification-loop

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.80). The skill runs grep searches for secret patterns like "sk-" and "api_key" and asks the agent to produce a verification report and diff/security findings without instructing redaction, so the agent may capture and output actual secret values verbatim (exfiltration risk).