1. Identify the desired Feishu/Lark action (e.g., create Bitable, send message).
2. Determine the appropriate `mcp__lark-mcp__` tool based on the action.
3. Gather necessary parameters (app_token, table_id, chat_id, etc.) from the user or context.
4. Format the request data according to the tool's specifications (JSON string for messages, array for filter values).
5. Handle authentication using `useUAT` (user vs. tenant token).
6. Execute the API call with the formatted request.
7. Return the result to the user.