tencent-survey

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

• Third-party content exposure detected (medium risk: 0.30). 该 skill 的运行时主要通过调用工具 get_survey / list_answers 在腾讯问卷服务端读取数据（包括问卷标题/DSL 与问卷回答文本），这些都是用户所提供的 survey_id/链接所指向的“第三方内容”，属于运行时会被 LLM 消费的自由文本来源。

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

• Potentially malicious external URL detected (high risk: 0.90). The skill invokes a remote MCP service that the agent calls at runtime (e.g., via mcporter) and that service URL "https://wj.qq.com/api/v2/mcp" can return an "instruction" used to tell the agent how to update/act, meaning fetched remote content can directly control agent instructions.