review-work

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.80). The prompt requires collecting and pasting full DIFFs and FILE_CONTENTS into spawned-agent prompts (and auto-collecting git outputs), which can force the LLM to handle or embed verbatim secrets from those files — an explicit redaction admonition exists but the prompt still instructs unredacted inclusion, creating a non-negligible exfiltration risk.