The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's primary function is to consume and process data from external, attacker-controllable sources such as Lark/Feishu chat messages, meeting notes, task descriptions, and whiteboard updates.
Ingestion points: Data enters the agent context through the lark-cli event consume command across various event domains (IM, Minutes, Task, VC, Board), as described in SKILL.md and the domain-specific reference files.
Boundary markers: The instructions do not specify the use of boundary markers or delimiters to isolate untrusted event content from the agent's instructions.
Capability inventory: The skill utilizes subprocess execution to run lark-cli and jq for event processing, providing a capability surface that could be targeted by instructions embedded in the event data.
Sanitization: There is no evidence of sanitization or filtering of the external content to remove potentially malicious instructions before the agent processes the data.
[COMMAND_EXECUTION]: The skill relies on executing shell commands via the lark-cli binary to interact with the Lark platform. While this is the intended functionality, it establishes a execution environment that must be carefully managed when processing the untrusted data ingestion described above.

lark-event