The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill facilitates the processing of asynchronous events from Lark interactive cards (card.action.trigger), which ingest untrusted input from users that could potentially influence agent behavior.
Ingestion points: Events are monitored via the lark-cli event consume command as documented in references/lark-im-card-action-reply.md, capturing values like form_value, input_value, and action_value.
Boundary markers: The skill explicitly instructs the agent to prompt the user for confirmation before activating the background listener and requires articulating visual changes before performing automated card updates.
Capability inventory: The skill possesses messaging and card update capabilities through lark-cli operations such as POST /open-apis/interactive/v1/card/update.
Sanitization: Content processing relies on the agent's logic to validate and reconstruct card structures from the event data, guided by the provided Card 2.0 schema and styling rules.

lark-im