feishu-multi-agent

What This Skill Does

The feishu-multi-agent skill provides a production-grade blueprint for orchestrating complex AI agent teams using the OpenClaw framework. It enables a hierarchical structure where a lead agent (the "Foreman" or "包工头") dispatches tasks to multiple subordinate agents. By leveraging Feishu (Lark) as the primary communication gateway, this skill establishes a file-driven task queue system. It uses custom workspace directory structures to ensure modularity, allowing agents to operate in isolation while maintaining a centralized command-and-control workflow. The architecture is specifically designed to bypass limitations in standard agent environments by utilizing cron-based heartbeat triggers and robust file-system orchestration.

Installation

To install this skill, you must first ensure your OpenClaw environment is configured with multiple Feishu app credentials. Add the skill to your foreman agent's SOUL.md configuration:

1. Use the install command: clawhub install openclaw/skills/skills/evan966890/feishu-multi-agent.
2. Create dedicated workspace directories for each subordinate agent (e.g., workspace-buyer, workspace-kb, workspace-gf) within your home directory.
3. Initialize the task structure by running mkdir -p ~/.openclaw/workspace-*/tasks/{pending,done}.
4. Update your OpenClaw gateway configuration to bind specific Feishu accounts to corresponding agent IDs.
5. Restart the OpenClaw gateway service to apply the configuration changes.

Use Cases

• Automated Business Workflows: Automating tasks like publishing articles, processing sales orders, or managing game industry production pipelines where agents need to work asynchronously.
• Multi-Agent Collaboration: Deploying specialized agents that communicate through file-based queues, ensuring that the main foreman can track progress without being blocked by individual task latencies.
• Continuous Background Automation: Utilizing cron job integration to ensure agents remain active 24/7, processing tasks as soon as they appear in their pending/ directories.

Example Prompts

1. "Check the status of pending tasks in the buyer and kb workspaces, and report any failures to the main dashboard."
2. "Create a new high-priority research task in the kb/tasks/pending folder regarding the latest AI market trends."
3. "Summarize the completed tasks from the done folders of all subordinate agents and draft a weekly progress report."

Tips & Limitations

• Workspace Isolation: Use bash commands for cross-workspace interactions, as the built-in write tools are restricted to individual agent workspaces.
• Cron Integration: For sustained activity, configure high-frequency cron jobs that monitor for new files, ensuring a seamless flow from pending to done states.
• Scaling: When adding more than 5 agents, monitor the Feishu WebSocket connection limits and consider rotating app credentials to avoid rate limiting. Always verify permissions for each bot application in the Feishu Developer Console before deployment.