seo-article-review

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

• Third-party content exposure detected (medium risk: 0.65). SKILL 运行时会把用户提供的 materialText/materialFile/materialUrl 作为“文章稿件/正文/链接内容”进入 contextComposer；若 materialUrl 指向公开网页或 materialText/上传文件包含他人撰写的内容，则该外部自由文本会被 LLM 读取并进入上下文（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). scripts/run.py sends a POST at runtime to the configured AISKILLS_BASE_URL (default https://ai-skills.ai/api/execute), which is a required remote call that executes the skill and thus allows external content to control the agent's execution/responses.