What problem does it solve? Handling Feishu (Lark) approvals manually through the web UI is slow and hard to automate. This Skill lets an AI agent query approval to-do/done lists, view instance details, and execute approval actions (approve, reject, transfer, rollback, remind, add-sign, cancel, cc) or initiate native approval instances directly through lark-cli. ## Core Features & Use Cases - Approval Task Processing: Query pending/completed approval tasks and execute actions like approve, reject, transfer, add-sign, rollback, and remind using instance_code and task_id pairs. - Approval Initiation Workflow: Search launchable approval definitions, inspect form controls and node requirements, then create native approval instances with correctly structured form values. - Instance Management: View instance details, cancel initiated approvals, and append cc recipients to running instances. - Use Case: A user asks "approve all my pending leave requests" — the agent queries the to-do list, retrieves instance_code and task_id pairs, and executes approvals with comments, handling error code 1395001 with a defined decision tree instead of blind retries. ## Quick Start Ask the agent to list my pending Feishu approvals and approve the leave request from yesterday with the comment "agreed".