The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is designed to interact with the Lark platform by executing commands through the lark-cli binary. All identified command patterns (e.g., lark-cli calendar events create, lark-cli calendar +agenda) are legitimate and directly related to the skill's core functionality for managing schedules and meeting rooms.
[PROMPT_INJECTION]: The skill provides an attack surface for indirect prompt injection by processing untrusted user data and interpolating it into command-line arguments.
Ingestion points: User-supplied text for meeting titles (--summary), descriptions (--description), and participant IDs (--attendee-ids) in references/lark-calendar-create.md and references/lark-calendar-room-find.md.
Boundary markers: Present. The skill mandates specific formatting for inputs, such as ISO 8601 for timestamps and standard ID prefixes (e.g., ou_, oc_, omm_), which serves as a functional delimiter.
Capability inventory: The skill utilizes subprocess calls via lark-cli to perform read and write operations on calendar data, including searching, creating, and deleting events.
Sanitization: Absent. The instructions do not explicitly require the agent to sanitize or escape user-provided strings before they are passed to the CLI tool, relying on the platform's underlying tool execution safety.

lark-calendar