product-research

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

• Third-party content exposure detected (high risk: 0.90). The skill explicitly fetches public product data and user-generated reviews via Sorftime MCP API (see scripts/get_reviews.py calling client.get_product_reviews and scripts/collect_data.py / references/api-*.md using category_report, keyword_detail, product_reviews), and those raw Top100 titles and competitor_reviews.json are then fed to LLM prompts in SKILL.md (Steps 2–4 VOC/交叉分析) to drive analysis and go/no-go decisions, so untrusted third-party content can directly influence agent behavior.