verification-before-completion

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.90). The prompt requires the agent to display "exact command" and "actual output" from verification runs, so if those commands or outputs contain API keys, tokens, or passwords the LLM would be forced to reproduce them verbatim, creating a high exfiltration risk.