cross-agent-collab

What This Skill Does

The cross-agent-collab skill provides a standardized framework for OpenClaw agents to interact with other AI agents effectively. It establishes essential protocols for communication, file sharing, and task coordination to prevent common pitfalls such as infinite feedback loops, cross-system incompatibility, and security vulnerabilities. By implementing these guidelines, agents can ensure efficient, professional, and safe interactions in shared environments like Discord channels or collaborative workspaces.

Installation

To integrate this skill into your environment, use the OpenClaw package manager: clawhub install openclaw/skills/skills/xiwan/cross-agent-collab

Use Cases

• Multi-Agent Coordination: Working alongside other bots (like mexico) to divide research or technical tasks.
• Cross-System Communication: Exchanging information between agents running on different frameworks (e.g., Clawdbot vs. OpenClaw).
• Joint Task Execution: Collaborating on shared document creation, joint reviews, or combined data analysis projects.
• Channel Etiquette: Maintaining a clean communication flow in shared spaces to avoid cluttering human-readable feeds.

Example Prompts

1. "@mexico, let's start the joint review for the project documentation. I have attached the draft here; please check sections 1-3, and I will handle 4-6."
2. "mexico, I need to compare our dataset results. Please share your current analysis file so we can consolidate our findings for the owner."
3. "agent 协作: Can you review this script? I've attached the latest version for your feedback before we submit it to the owner."

Tips & Limitations

• Strict NO_REPLY Policy: Silence is encouraged. Do not send 'received' or 'ok' messages unless specifically requested to confirm receipt. This prevents bot-to-bot echo loops.
• System Boundary Awareness: Never attempt to troubleshoot or provide configuration advice for an agent's underlying system that you do not host. If a system error occurs, flag it to the human owner immediately.
• File Sharing Protocol: Always use the message(action=send, ...) format with valid attachments. Never reference local file paths that the other agent cannot access.
• Data Privacy: Never share API keys, credentials, or personal information about your owner. Treat other agents as external collaborators—trust but verify information directly with them.