What problem does it solve? Building event-driven automations on Lark (Feishu) normally requires standing up a webhook server, handling verification challenges, and parsing raw payloads. This Skill lets you listen to Lark events in real time over a WebSocket long connection and receive clean NDJSON directly on stdout, ready for shell pipelines and AI agents. ## Core Features & Use Cases - Real-time event subscription: Connect via WebSocket long connection and receive messages, contact changes, calendar updates, approvals, tasks, and drive events as NDJSON. - Agent-friendly compact output: The --compact mode flattens payloads, parses double-encoded message content, and strips noise fields so LLM pipelines can consume events directly. - Regex routing and file output: Route different event types to different directories with --route, or write each event to a file with --output-dir. - Use Case: Pipe incoming IM messages into Claude to generate replies, then respond as the bot via lark-cli api, creating a fully automated Lark chatbot loop. ## Quick Start Ask the agent to subscribe to incoming Lark messages with lark-cli event +subscribe in compact mode and reply to each message automatically.