wechat-article-generator

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

• Third-party content exposure detected (high risk: 0.75). 当 enable_research=true 且 input_type=keyword 时，工作流会对关键词执行运行时 WebSearch 获取外部网页文本并将其用于生成文章，从而把“公共网页/搜索结果摘要”这类外部自由文本喂入LLM上下文。

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly uses WebFetch/WebSearch at runtime to read user-supplied URLs (e.g., the README examples like https://react.dev/learn and https://docs.docker.com/get-started/) and injects that fetched content into the model context to drive article generation, so remote URLs can directly control prompts/outputs.