The Agent Skills Directory

Indirect Prompt Injection Surface: The skill processes untrusted external data by scanning email threads and CRM deal information. If an external email contains text designed to override instructions, it could potentially influence the agent's behavior during the prioritization process. Ingestion occurs in Step 1 where email threads are pulled. There are no explicit boundary markers or instructions to ignore embedded commands within the data.
Command Execution Capability: The skill is granted access to the Bash tool to perform its logic. When an agent has command-line capabilities and processes untrusted input (such as email content), there is a theoretical risk of command injection if the agent attempts to execute text found within an email as a command. The Bash tool is listed in the allowed-tools configuration.
Sensitive Data Access: The skill interacts with highly sensitive data sources, including Mail and HubSpot CRM records. While this access is essential for lead ranking and providing context, it represents a significant data surface that relies on the secure handling of the underlying platform integrations.

call-list