source-reading-analyst

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.80). The prompt mandates quoting "精彩代码片段" (5–15 lines) and file-evidence verbatim without guidance to redact secrets, so if the codebase contains API keys/passwords the model would be expected to reproduce them, creating an exfiltration risk.