api-doc-integration-risk-review

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

• Third-party content exposure detected (medium risk: 0.65). 该技能在运行时会把用户提供的 apiGuideText/apiGuideFile/apiGuideLink（尤其是公开可访问的 URL 或上传的外部文档内容）作为可读文本进入 LLM 上下文；其中 apiGuideLink 对应“运行时抓取的公共网页/开发者文档”属于外部来源，存在间接提示注入风险。

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

• Potentially malicious external URL detected (high risk: 0.90). The provided runner calls the external service at https://ai-skills.ai (POST to https://ai-skills.ai/api/execute) at runtime to execute the skill and return LLM-generated instructions/results that directly control agent behavior, which is a required external execution dependency.