The Agent Skills Directory

[COMMAND_EXECUTION]: The skill interacts with the system by executing shell commands via the lark-cli binary. These commands are used to create, update, delete, and list tasks, as well as manage task members and reminders (e.g., lark-cli task +create, lark-cli task +update).
[INDIRECT_PROMPT_INJECTION]: The skill processes external, user-controlled data retrieved from the Lark API, which could potentially contain malicious instructions intended to influence the agent's behavior.
Ingestion points: External task titles, descriptions, and comments enter the agent's context through commands like +get-my-tasks and native list or get operations defined in SKILL.md.
Boundary markers: The instructions do not define specific technical boundary markers (such as XML tags or JSON schemas) for all retrieved text, but they do include a mandatory safety rule requiring the agent to confirm the user's intent before performing any write operations.
Capability inventory: The skill has the capability to perform write operations (create, patch, delete) on tasks, tasklists, and comments, and to manage workspace members.
Sanitization: No explicit programmatic sanitization is defined within the skill files for the retrieved task content; the skill relies on the agent's internal safety filters and the explicit instruction to seek user confirmation.

lark-task