What problem does it solve? Building Lark/Feishu bots and integrations that react to real-time events (messages, approvals, meetings, tasks) normally requires standing up webhook servers and handling push infrastructure. This Skill lets an AI agent subscribe to and consume Lark events directly as NDJSON streams through the lark-cli event command, with a defined subprocess contract for readiness, bounded runs, and graceful shutdown. ## Core Features & Use Cases - Event discovery and schema inspection: List all subscribable EventKeys by domain (im, approval, vc, task, minutes, board, application) and inspect each key's output schema and jq root path before consuming. - Streaming consumption with filtering: Consume events as NDJSON on stdout with jq-based filtering/projection, plus bounded runs via --max-events and --timeout. - Subprocess contract for AI agents: A stderr ready marker, stdin-EOF graceful shutdown, structured JSON error envelopes, and documented exit codes make it safe to run as a managed subprocess. - Use Case: Build a bot that listens to im.message.receive_v1 as the bot identity, filters for text messages in group chats, and reacts to each incoming message in real time. ## Quick Start Ask the AI to list available Lark event keys with lark-cli event list, then start consuming im.message.receive_v1 as the bot identity to stream incoming messages.