The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes text from meeting transcripts and document content, which are external data sources that could potentially contain malicious instructions.
Ingestion points: Transcripts are fetched via lark-cli note +transcript and saved to the filesystem, while document contents are read using docs +fetch (as described in SKILL.md and lark-note-detail.md).
Boundary markers (absent): The instructions do not define clear delimiters or specific directions for the agent to isolate or ignore instructions embedded within the fetched content.
Capability inventory: The skill allows for shell command execution through lark-cli and file writing to the local ./notes/ directory.
Sanitization (absent): There is no mention of filtering or sanitizing the retrieved content before it is displayed or saved.
[COMMAND_EXECUTION]: The skill uses a required binary to perform its tasks.
Evidence: The skill executes subcommands of the lark-cli binary, such as note +detail and note +transcript, which is standard behavior for this vendor-authored tool.

lark-note