patent-mining-agent

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The prompt mandates recording the user's CP-0 confirmation fields (including "user_extra") and later writing them verbatim into the final report, which forces the LLM to output any user-supplied values — including secrets — exactly as given.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.30). Skill 通过 patent.search/web.search 检索并在 Block 1 的“专利扫描”与后续 Block 3 的“现有技术清单/检索”中读取检索到的公开网页与专利文本，而这些检索内容可能包含外部作者注入的自由文本。