sn-dimension-research

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

• Third-party content exposure detected (high risk: 1.00). The skill explicitly routes to general web/page-opening and social search skills (e.g., "通用网页检索与页面打开能力" and sn-search-social-cn / sn-search-social-en with examples like B站、知乎、抖音、Reddit、X、YouTube) and requires the agent to read and act on those public, user-generated sources as part of its required search-and-decision workflow, so untrusted third-party content can influence behavior.