ucoz-shop-optimizer-skill

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

• Third-party content exposure detected (high risk: 0.75). The skill’s required runtime workflow ingests shop content via uCoz Online Shop API calls (e.g., GET /shop/request?page=categories and GET /shop/request?page=viewgoods&id={id}), where fields like category/product descriptions and HTML can contain outsider-authored free text (e.g., user-written product/category copy) that the agent then places into the LLM context for analysis/planning.

MEDIUM W009: Direct money access capability detected (payment gateways, crypto, banking).

• Direct money access detected (high risk: 1.00). The skill exposes explicit shop payment and order APIs that can perform financial actions: e.g., POST /shop/checkout/ (submit order), PUT /shop/invoices/change (change payment method and restart payment), PUT /shop/invoices/status and other order-modifying endpoints (change status, courier, delivery cost via PUT /shop/order/delivery_tax), currency-rate update (POST /shop/setcurrrate), and basket operations (POST/PUT/DELETE /shop/basket/) which can create/modify/clear pending purchases. These are specific, non-generic endpoints capable of initiating or altering payment/transaction flows (i.e., sending transactions), even though the skill adds safeguards and requires explicit approval for such actions. Because the skill includes direct checkout/payment/order-modification endpoints, it grants direct financial execution capability.