The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted Markdown content retrieved from remote Lark Drive or Wiki via +fetch, +diff, and +patch operations.
Ingestion points: Remote Markdown content is ingested into the agent's context through lark-cli markdown +fetch (returns file content), lark-cli markdown +diff (returns differences between versions), and lark-cli markdown +patch (downloads and replaces content).
Boundary markers: There are no explicit instructions or delimiters defined to help the agent distinguish between the retrieved file content (data) and its own instructions.
Capability inventory: The skill utilizes the lark-cli binary, which has the capability to write to the local filesystem (via --output in +fetch) and perform remote write operations in the Lark environment (+create, +overwrite, +patch).
Sanitization: The instructions do not specify any sanitization, filtering, or validation of the remote Markdown content before it is processed or displayed by the agent.

lark-markdown