The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process external data that could contain malicious instructions targeting the agent.
Ingestion points: Untrusted data enters the agent context through email content parameters (e.g., email__config__html__content) in the SENDGRID_CREATE_SINGLE_SEND tool and contact data in SENDGRID_IMPORT_CONTACTS and SENDGRID_ADD_OR_UPDATE_A_CONTACT.
Boundary markers: The skill instructions do not specify any delimiters or explicit warnings to the agent to disregard instructions embedded within the email bodies or CSV data.
Capability inventory: The skill has broad capabilities to send emails account-wide, manage sender identities, and manipulate contact databases via the SendGrid API.
Sanitization: No sanitization or validation steps are described for external content before it is processed by the tools.
[EXTERNAL_DOWNLOADS]: The skill requires the configuration of a third-party MCP server, which acts as a gateway for the described tools.
Evidence: The setup section instructs the user to add https://rube.app/mcp as an MCP server. This routes tool definitions and potentially sensitive operational data through external infrastructure not controlled by the primary service provider.

sendgrid-automation