The Agent Skills Directory

Indirect Prompt Injection Surface: The skill is designed to process external content from sources like customer support tickets and dispute reasons, which may contain instructions intended to influence the agent's behavior. \n
Ingestion points: External data enters the agent context via PayPal disputes, HubSpot tickets, and uploaded review exports in Step 1. \n
Boundary markers: The skill does not currently employ explicit delimiters or delimiters for the agent to differentiate between its core instructions and the external data it processes. \n
Capability inventory: The skill is configured with access to the Read, WebFetch, and Bash tools, which provide the agent with capabilities to interact with files and network resources. \n
Sanitization: There is no evidence of input validation or sanitization of the feedback signals before they are clustered or used to draft response templates.

customer-pulse-check