The Agent Skills Directory

[PROMPT_INJECTION]: The skill identifies a surface for indirect prompt injection as it processes message IDs originating from external email content (SKILL.md). It mitigates this risk by including a clear disclaimer instructing the agent to treat such data as untrusted and to avoid following any embedded instructions. The skill also limits its capability by validating that inputs conform to specific message ID formats, preventing schema confusion.
[COMMAND_EXECUTION]: The skill utilizes curl within a Bash shell to perform POST requests to a specified webhook URL. This command execution is the primary functional requirement for the skill to interact with the n8n automation server and is restricted to batch processing of message IDs.
[DATA_EXFILTRATION]: The skill transmits message IDs to an external n8n host. This operation is a core functional requirement and is documented with placeholders for the host and workflow ID, along with instructions to secure the endpoint using a shared secret stored in an environment variable.

mark-read