Multi-Agent Intercom

Version 1.1.6 of multi-agent-intercom - No code or documentation changes detected. - Version or metadata update only; functionality remains the same.

- Switched initialization command from install to setup; users must now run python <SKILL_DIR>/scripts/intercom.py setup to generate the SOP. - Setup script prints the SOP; users are now required to manually paste it into each agent's AGENTS.md. - Added clearer and expanded usage instructions, command reference, and security notes. - Removed all mentions of auto-writing files during setup for enhanced safety and transparency. - Updated feature list to highlight i18n, anti-loop, and improved OS compatibility.

No changes detected in this version. - Version 1.1.2 was published with no file modifications. - No updates or feature changes from previous release.

Multi-Agent Intercom 🗣️ A powerful peer-to-peer communication skill for OpenClaw. This skill enables independent OpenClaw agents (e.g., zz, dev, rc) to securely send messages and wake up each other across isolated workspaces. It acts as an "intercom system" for your Multi-Agent setup. 🌟 Why is this needed? By design, OpenClaw agents run in isolated databases and sandboxes. The built-in sessions_send tool can only send messages within an agent's own subagent hierarchy. If Agent A tries to use sessions_send to message Agent B, it will result in a "Session not found" error. Multi-Agent Intercom solves this by leveraging the native openclaw agent CLI mechanism to safely bridge these isolated environments, allowing agents to act as equal peers in a decentralized company. ✨ Features Auto-Discovery: Agents can automatically list all other agents configured on the current Gateway. Cross-Boundary Messaging: Send instructions or notifications to any agent on the same system. Internationalization (i18n): Automatically detects the OS language and injects SOPs/Caller IDs in either English or Chinese. Absolute Stealth (Cross-Platform): Executes asynchronously in the background. Zero terminal popups on Windows, fully detached on macOS/Linux. Immediate Wake-up: Bypasses sleep states and forces the target agent to process the message immediately. Persistent Context: Targets the main session by default, ensuring long-term context is preserved. 📦 Installation & Setup If you downloaded this from ClawHub or GitHub: Extract the folder into your ~/.openclaw/skills/ directory. Ensure the OpenClaw CLI (openclaw) is accessible in your system's PATH. The latest version fully complies with ClawHub review standards and has been published at https://clawhub.ai/zhangxue1985122219/multi-agent-intercom. The latest version is not the most powerful—it was adjusted to meet community guidelines. If you can read the code, the previous version is actually the most powerful; continue using that one. One-Time Initialization: Ask any of your agents to install the intercom protocol using natural language: "请安装 multi-agent-intercom 技能" (Or: "Please run the installation script for the multi-agent-intercom skill.") The agent will autonomously read the SKILL.md and run python scripts/intercom.py install. This safely injects the communication SOP into the AGENTS.md of every agent on your system. After installation, type /new in your chat to reload their brains with the new protocol. 🚀 How Agents Use It Agents can use this skill autonomously when they determine they need to contact another agent. The scripts are entirely silent and cross-platform (no popup terminal windows on Windows or Mac). To list available agents: python scripts/intercom.py list To send a message: The syntax requires the sender to identify themselves so the receiver gets a proper "Caller ID". python scripts/intercom.py send <TARGET_ID> <SENDER_ID> "Your message here" Example (Agent 'zz' messaging 'dev'): python scripts/intercom.py send dev zz "Hi dev, the user wants us to update our avatars. Please acknowledge." When 'dev' receives this, the system automatically formats it with a clean, system-compatible tag: [来自智能体 dev] Your message here Thanks to the injected AGENTS.md SOP, 'dev' instantly knows how to reply using the exact same tool. 🛡️ Requirements OpenClaw >= 2026.x Python 3.x

multi-agent-intercom v1.1.2 - Updated the skill description for improved clarity and to highlight zero console popups and multiline support. - Clarified setup instructions: the skill will no longer attempt or offer to run the setup script; users must run it manually. - Refined 'when' and 'examples' triggers for higher accuracy and relevance. - Removed installation-related trigger phrases to avoid confusion. - Minor documentation cleanups for better user guidance.

No changes detected in this version (1.1.1): - No file changes or updates since the previous release. - Functionality and documentation remain unchanged.

- Introduced a new security-compliant setup process requiring user consent and manual SOP installation in AGENTS.md for each agent. - Replaced the installation step with a "setup" script: use python <SKILL_DIR>/scripts/intercom.py setup to generate SOP instructions. - Updated instructions to highlight that agents' files cannot be modified automatically; users must manually apply changes. - Streamlined documentation and examples for clarity. - Emphasized asynchronous, zero-intrusive operation and improved compliance messaging.

**A powerful peer-to-peer communication skill for OpenClaw.** This skill enables independent OpenClaw agents (e.g., `zz`, `dev`, `rc`) to securely send messages and wake up each other across isolated workspaces. It acts as an "intercom system" for your Multi-Agent setup ## ✨ Features - **Auto-Discovery**: Agents can automatically list all other agents configured on the current Gateway. - **Cross-Boundary Messaging**: Send instructions or notifications to any agent on the same system. - **Internationalization (i18n)**: Automatically detects the OS language and injects SOPs/Caller IDs in either English or Chinese. - **Absolute Stealth (Cross-Platform)**: Executes asynchronously in the background. Zero terminal popups on Windows, fully detached on macOS/Linux. - **Immediate Wake-up**: Bypasses sleep states and forces the target agent to process the message immediately. - **Persistent Context**: Targets the `main` session by default, ensuring long-term context is preserved.