The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on executing the lark-cli binary with various subcommands such as note +detail, note +transcript, and vc +detail. These commands incorporate identifiers like note_id and meeting_id which are obtained from user input or previous command outputs.
[EXTERNAL_DOWNLOADS]: The skill requires the lark-cli binary to be present in the environment as specified in the metadata requires.bins. This tool is a vendor-provided utility from Larksuite.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection as it retrieves and processes meeting transcripts and document content (verbatim_doc_token, unified_transcript.md). These sources contain arbitrary text generated by users or meeting participants.
Ingestion points: Meeting transcripts are read via lark-cli note +transcript and documentation via lark-cli docs +fetch (in SKILL.md and references/lark-note-transcript.md).
Boundary markers: The skill does not explicitly define prompt delimiters for the ingested data within its instructions, though it emphasizes strict routing and identity continuity.
Capability inventory: The agent has access to shell execution (lark-cli) and local file system writes (./notes/{note_id}/).
Sanitization: No explicit sanitization of the transcript text is mentioned in the instruction files; it is treated as raw text output.

lark-note