langchain-rag

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

• Third-party content exposure detected (medium risk: 0.30). 该技能的运行路径是“查询→embeddings→vectorStore检索→把检索到的document.pageContent拼入system上下文→LLM生成”，因此如果向量库/检索到的文档内容来自外部或用户可提交的文本，就会把该外部自由文本间接暴露给LLM。

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The CheerioWebBaseLoader example fetches external content at runtime from "https://docs.langchain.com/oss/javascript/langchain/agents" which is then loaded into documents and later injected into the LLM prompt context, so this URL can directly control agent instructions.