The Agent Skills Directory

[DATA_EXFILTRATION]: The skill transmits user-provided sensitive data to an external API endpoint.- The createOrder and validateOrder tools send user email addresses and phone numbers to https://api.cryptorefills.com/mcp/http.- This data transmission is required for order fulfillment and mobile credit delivery, representing a legitimate but significant data flow to a non-whitelisted domain.- [INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection due to processing external data.- Ingestion points: Untrusted data enters the agent context via the output of the searchProducts, listBrands, listProductsForCountry, and getOrderStatus tools (e.g., brand names, product descriptions, and order messages).- Boundary markers: None identified in the prompt instructions to delineate or warn the agent about potential instructions embedded in API-returned text.- Capability inventory: The skill has the capability to perform external state changes via the createOrder and validateOrder tools, allowing a potential injection to trigger unauthorized or unintended purchase attempts.- Sanitization: No specific sanitization, escaping, or filtering of API-returned content is defined in the instructions before the information is presented to or processed by the agent.

cryptorefills-buy