weiyun

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

• Third-party content exposure detected (medium risk: 0.30). 源自 Weiyun 的 MCP 更新检查流程会从服务端返回的 release_note / instruction 文本中让客户端（AI Agent）读取并据此执行更新指令：mcporter call https://www.weiyun.com/api/v3/mcpserver check_skill_update → 解析 JSON 字段。

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

• Potentially malicious external URL detected (high risk: 0.90). High confidence: the skill requires a daily runtime call to https://www.weiyun.com/api/v3/mcpserver (used by check_skill_update and as the default MCP endpoint), and the update API returns an "instruction" field which the skill tells the agent to "follow" — meaning remote content can directly control agent behavior.