The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses lark-cli and whiteboard-cli to perform operations on the Lark platform. These include document searches, content fetching, and updating document blocks. These are legitimate but powerful operations performed via shell commands.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection. It retrieves content from Lark documents using docs +fetch. This content, which could be controlled by an external party, is then processed by the agent. Without boundary markers or sanitization, the agent might interpret text within the document as direct instructions.
Ingestion points: Document content retrieved in references/lark-doc-fetch.md via the --doc parameter.
Boundary markers: None identified in the skill instructions to separate document content from agent instructions.
Capability inventory: Extensive capabilities through lark-cli, including content modification (docs +update) and permission management (granting full_access in lark-doc-create.md).
Sanitization: No sanitization process is described for the fetched Markdown content before processing.

lark-doc