category-selection

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

• Third-party content exposure detected (high risk: 1.00). The skill explicitly calls external Sorftime MCP tools (e.g., category_report, product_detail, category_keywords shown in SKILL.md and implemented in scripts/analyze_category.py) to fetch public Amazon/TikTok/1688 data and parses textual fields (titles, descriptions, reviews, keywords) that are then used to compute scores and recommendations, so untrusted third‑party content is ingested and can materially influence agent decisions.