Agent Mailbox Light
Provides a lightweight, cross-workspace internal mail system for temporary coordination among multiple AI agents.
Install & Use
Copy this prompt and send it to your AI assistant (Claude / Cursor / TRAE / Codex / WorkBuddy etc.) to auto-install:
Help me install this AI Skill: Agent Mailbox Light. It is used for: Provides a lightweight, cross-workspace internal mail system for temporary coordination among multiple AI agents. Full Skill content: https://321skill.com/skills/agent-mailbox-light-x-9/raw/index.md Read that page and install it.
The prompt includes a link to the full Skill content. You can also view the full content.
This Skill addresses the lack of a lightweight, low-noise, inter-session coordination channel between multiple AI agent workspaces. In practical development, developers often need different agents (e.g., those responsible for frontend, backend, or testing) to pass temporary prompts or context, but they don't want to write this information into long-term memory or rely on complex message buses. Agent Mailbox Light offers a simple, file system-based mailbox that allows information to flow between workspaces.
Usage is straightforward: simply initialize a .agent-mailbox folder in the root directory of each workspace, containing inbox, keep, and archive subdirectories. You can then use scripts to send mail files to matching workspace patterns (e.g., $HOME/.openclaw/workspace*). The receiving agent reads the latest mail in its inbox upon startup or at task entry points, processes it, and then moves the mail to keep or archive.
It is ideal for teams or individuals who need to coordinate multiple parallel agent tasks while maintaining independent, loosely coupled workspaces. This is especially useful for those already using frameworks like OpenClaw, where temporary, prompt-based information needs to be passed between different sessions (e.g., coding, testing, deployment agents)—such as synchronizing project context, coordinating multi-agent tasks, or communicating temporary configuration change alerts.
It is recommended to integrate mailbox checks into the agent's baseline startup routine, for example, by reading the latest 1-3 mails each time it starts. Note that this is primarily a "best-effort" broadcast system: it does not guarantee delivery, has no push notifications, and mails should be treated as "suggestive prompts" for the local agent to evaluate, not as mandatory commands. It is also limited to the local file system and does not support cloud sync or remote delivery.
Key Features
Unlike message queues (e.g., RabbitMQ) that require a central registry or guaranteed delivery, it adopts a minimalist, registration-free, best-effort broadcast model, relying solely on the local file system to achieve zero-dependency, cross-workspace temporary coordination.
Limitations
Only supports the local file system; not suitable for cloud environments or scenarios requiring network synchronization. Failed sends are not retried.
FAQ
Are failed mail sends retried?
No. This is a best-effort system. If writing to the target workspace's inbox fails, the sender simply skips it; there is no retry queue or failure notification.
Will mail content consume a large number of tokens?
The design encourages brief mails. The receiving agent typically only reads the latest 1-3 mails and compresses them into a concise summary to optimize token consumption.
Installation guide for AI assistants
If your AI coding assistant (Claude Code, Cursor, TRAE etc.) can see this page, send it this message to auto-install:
Visit https://321skill.com/skills/agent-mailbox-light-x-9/raw/index.md to read the original Skill definition (Markdown format) for Agent Mailbox Light, and install it according to the instructions.
Raw Markdown URL for AI: /skills/agent-mailbox-light-x-9/raw/index.md