ai-podcast

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

• Third-party content exposure detected (high risk: 0.75). 该技能的运行时上下文由用户通过 JSON 参数提供的 materialText/materialFile/materialUrl 组成；其中 materialUrl 对应的公开网页内容会在技能执行时被读取并转为可读文本进入 LLM（属于“公共 web 内容在运行时抓取/浏览器提取”这一类外部自由文本）。

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

• Potentially malicious external URL detected (high risk: 0.90). The provided runner script performs runtime calls to the remote skill execution endpoint at https://ai-skills.ai (POST to /api/execute using AISKILLS_API_KEY), meaning execution and prompts are controlled by that external service.