meta-cognition-parallel

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.90). The skill explicitly embeds the raw $ARGUMENTS (the user's question) into subagent prompts and the final "Query: [User's question]" output, so any API keys/passwords included in the user's input would be forwarded and echoed verbatim—creating a high exfiltration risk.