a1-yandex-kit-catalog-doctor

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

• Third-party content exposure detected (medium risk: 0.30). In the required runtime workflow, the agent performs read-only “catalog audit” calls via the Yandex KIT MCP server (e.g., list_products/get_product, list_variants/get_variant, list_collections/get_collection, etc.) and then classifies evidence from the returned catalog objects, which can include outsider-authored free text fields (like product/category/media metadata) without first selecting a specific trusted item.