zsxq

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

• Third-party content exposure detected (medium risk: 0.30). 在 references/scenarios/daily-patrol.md 的运行路径中会先用 group +topics / topic +detail（以及必要时 get_topic_comments）读取星球帖子正文与评论文本后由模型做语义判断与“处理建议”生成，因而会把第三方成员发布的自由文本喂给模型；攻击者只需在星球内发布毒化内容即可影响输出。

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

• Potentially malicious external URL detected (high risk: 0.90). The skill includes runtime instructions that fetch and execute remote npm packages (e.g., npm install -g zsxq-cli, npx http-server, npx qrcode, and scripts that require puppeteer), and the package-lock shows a resolved package URL https://registry.npmjs.org/puppeteer/-/puppeteer-22.15.0.tgz which is remote code fetched and executed at runtime.