The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from email sources, creating a potential surface for indirect prompt injection. Ingestion points: Untrusted data enters the agent's context through tools like get_email, get_email_context, and get_thread (SKILL.md, tools.md). Boundary markers: The instructions explicitly command the agent to treat email bodies, headers, and attachments as untrusted data and to ignore any embedded instructions (security.md). Capability inventory: The skill has capabilities to move, delete, and modify inbox states, as well as manage folders and labels (tools.md). Sanitization: The skill implements safety measures such as agent_safe_content flags and scan status requirements to sanitize incoming data (SKILL.md, tools.md).
[EXTERNAL_DOWNLOADS]: The skill connects to an external MCP server at https://console.mermail.app/mcp to perform mailbox operations.
[COMMAND_EXECUTION]: The skill facilitates the execution of various mailbox management commands through the Mermail MCP server.

mermail-manage-inbox