capture-feedback

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill explicitly instructs the agent to capture and emit the user's correction verbatim (including running a script with the feedback as a command-line argument and writing it into a local JSON), so any secrets or credentials supplied by the user would be reproduced verbatim in the agent's output, creating high exfiltration risk.