The Agent Skills Directory

[PROMPT_INJECTION]: The skill creates an indirect prompt injection attack surface by consuming data from the Lark event stream, such as incoming messages or calendar updates, which originate from potentially untrusted external users. \n
Ingestion points: Lark WebSocket event stream accessed via lark-cli event +subscribe in references/lark-event-subscribe.md. \n
Boundary markers: None identified in the provided instructions or examples. \n
Capability inventory: The skill can write events to the local file system (--output-dir) and the documentation provides examples of piping this data to other tools like claude for automated replies or lark-cli docs for updating documents. \n
Sanitization: No validation or sanitization of the event content is performed before processing or outputting the data.

lark-event