simplify

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.80). The prompt requires the agent to forward "the full diff" to parallel review agents and produce code fixes/outputs using that context, which would force the LLM to read and potentially echo any literal secrets contained in the diff (creating exfiltration risk) even though it doesn't explicitly ask for credentials.