Cross Agent Chat

What This Skill Does

The Cross Agent Chat skill, developed by ebly, provides a standardized and robust protocol for Inter-Agent Communication (IAC) within the OpenClaw ecosystem. Instead of relying on direct manual messaging, this skill acts as a bridge, allowing an AI agent to programmatically discover, initiate, and maintain conversations with other agents. It manages the session lifecycle through the sessions_send tool, ensuring that requests are properly addressed to target agents—whether they are specialized assistants, code evaluators, or administrative bots—while handling communication timeouts and session identification seamlessly.

Installation

To integrate this capability into your OpenClaw environment, execute the following command in your terminal: clawhub install openclaw/skills/skills/ebly/cross-agent-chat

Use Cases

This skill is essential for building multi-agent architectures where modularity is key. Common use cases include:

• Capability Discovery: Querying another agent to determine if it has access to specific tools (e.g., asking a weather bot for current forecasts).
• Task Delegation: Offloading specialized workloads, such as requesting a developer agent to generate a unit test or asking a summary bot to parse a long report.
• Complex Orchestration: Coordinating multi-step workflows where one agent acts as a controller, distributing sub-tasks to various specialized agents and aggregating the results.

Example Prompts

1. "Look up the active session for the 'Code Reviewer' agent and ask them to verify the security of the script I just generated."
2. "Find the assistant agent with the label 'Data Analyst' and send them this CSV data to perform a trend analysis."
3. "List all available sessions in the last 24 hours, identify the 'Internal Support' agent, and ask them why the server connection is timing out."

Tips & Limitations

To maximize the efficiency of your agentic workflows, always perform a sessions_list before attempting to send a message. This ensures you are targeting the correct sessionKey and confirms the recipient is currently active. Avoid guessing keys. Keep your requests concise and structured; since you are communicating with another machine, clarity is preferred over conversational filler. Note that the skill relies on the availability of the recipient agent—if an agent is offline or does not have the necessary permissions, the request may return a failure. Always include a sensible timeoutSeconds value for long-running processes to prevent hanging your main agent's execution thread.