defend

SUSPICIOUS. The skill's stated purpose is coherent as a gate/review workflow, and there are no installers, payloads, or explicit exfiltration endpoints in the provided content. However, it routes sensitive operational artifacts into an external model-review step labeled 'ask-claude' that was not verified from the supplied evidence, and one gate directly blocks or permits mechanical trading changes without a human in the loop. Risk is driven more by unspecified data flow and autonomous real-world impact than by confirmed malicious behavior.