The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes shell commands using lark-cli to perform document operations and uses python3 for parsing JSON output and running local utility scripts. Evidence is found in references/lark-doc-mention-link.md which includes a Python one-liner for block parsing and references to a local script scripts/lark-mention-link.py for document auditing.
[EXTERNAL_DOWNLOADS]: The skill facilitates the downloading of media and files from the Lark platform. references/lark-doc-media-download.md and references/lark-doc-media-preview.md describe shortcuts for retrieving document assets. Additionally, references/lark-doc-create.md specifies that the system automatically downloads images from URLs provided within markdown tags to upload them to the Lark document space.
[PROMPT_INJECTION]: The skill creates a surface for indirect prompt injection because it retrieves and processes content from external Lark documents which may contain untrusted data.
Ingestion points: Untrusted document content is ingested via docs +fetch in references/lark-doc-fetch.md and docs +search in references/lark-doc-search.md.
Boundary markers: The instructions do not define clear boundary markers or guidelines to ignore instructions embedded within the retrieved document content.
Capability inventory: The skill possesses the capability to modify document content and execute shell-based tools (lark-cli, python3).
Sanitization: There is no evidence of content sanitization or validation performed on the retrieved document data before it is presented to the agent.

lark-doc