steve-jobs-perspective

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

• Third-party content exposure detected (high risk: 1.00). Flagged: SKILL.md Step 2 explicitly requires using WebSearch/tools to fetch public product reviews and user feedback (e.g., "⚠️ 必须使用工具（WebSearch等）获取真实信息" and the Vision Pro WebSearch example), so the agent will ingest untrusted third‑party web content that the workflow uses to drive decisions.