langchain-document-loaders

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

• Third-party content exposure detected (high risk: 0.75). 该技能的运行时工作流会使用诸如 CheerioWebBaseLoader / PlaywrightWebBaseLoader / GithubRepoLoader 等加载器在运行时抓取或克隆外部网页/仓库内容，并将其提取为 pageContent（可读自由文本）喂入 LLM，从而引入“公共网页/外部仓库文本”这类 outsider 来源的间接提示注入风险。

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

• Potentially malicious external URL detected (high risk: 0.90). The skill examples use document loaders that fetch remote content at runtime (e.g., GithubRepoLoader loading https://github.com/langchain-ai/langchainjs and web loaders loading https://docs.langchain.com, https://news.ycombinator.com, https://example.com/page1, https://spa-app.com, https://blog.com/post), which are injected as documents and can directly influence the agent's prompts/context.