The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The documentation provides instructions to install the agentmail package from official Python and Node.js registries. These are official SDKs provided by the vendor for interacting with the AgentMail API.\n- [DATA_EXFILTRATION]: The skill demonstrates how to read local files to send as email attachments via the api.agentmail.to endpoint. This is a primary function of the SDK and is used for intended platform features.\n- [PROMPT_INJECTION]: This skill processes untrusted external data in the form of incoming emails, which presents a surface for indirect prompt injection.\n
Ingestion points: External email content enters the agent context via the messages.list API call, Webhooks, or WebSockets (documented in SKILL.md, references/webhooks.md, and references/websockets.md).\n
Boundary markers: The documentation recommends using the extracted_text field to isolate new message content from quoted conversation history.\n
Capability inventory: The skill has the capability to read local files, send outbound communications, and manage account resources (documented in SKILL.md).\n
Sanitization: The platform provides automated stripping of signatures and quoted text through the extracted_text property to help sanitize incoming content before processing.

agentmail-sdk