Agent Mailbox Light
A lightweight, cross-session ephemeral coordination mailbox for multi-agent workspaces.
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: A lightweight, cross-session ephemeral coordination mailbox for multi-agent workspaces. Full Skill content: https://321skill.com/skills/agent-mailbox-light-x-7/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 coordination channel between multiple AI agent workspaces. In practical development, developers often need to pass ephemeral prompts, warnings, or context information between different agents (e.g., code assistant, documentation assistant, testing assistant), but do not want to write this information into long-term memory or use complex message buses to avoid token waste and system coupling.
Usage is simple: you only need to create a .agent-mailbox folder and its subdirectories (inbox, keep, archive) in the agent's workspace directory. You can then use scripts to write mail files (e.g., 20260322100000--main--warn--cron-list-broken.md) into the target workspace's inbox directory. Upon startup or at task entry points, the receiving agent automatically checks and reads the latest mail in its own inbox, extracts key information, and moves processed mail to archive or mail to be kept to keep.
It is ideal for teams or individuals needing to coordinate multiple AI agents running in parallel (e.g., code generation, testing, deployment agents). This is especially true for those already using frameworks like OpenClaw and wishing to establish a simple, stateless, cross-session mechanism for passing prompt information between different workspaces. It is particularly useful for managing complex project workflows and scenarios requiring temporary instructions or warnings to be passed between agents at different stages.
It is recommended to adopt this tool uniformly when integrating agents into automated workflows (e.g., CI/CD pipelines, multi-step task decomposition). Note that it primarily provides a best-effort delivery model, with no push notifications, delivery guarantees, or retry queues. Therefore, it is not suitable for transmitting mission-critical instructions and should be used as an auxiliary contextual prompt system. Mail content should be kept concise, and recipients are encouraged to read only the latest 1-3 mails to optimize token consumption.
Key Features
Unlike complex message queues (e.g., RabbitMQ) that require a central registry or guaranteed delivery, it employs a decentralized, best-effort delivery model based on the local filesystem, requiring no additional services for minimalistic implementation. Compared to writing everything into long-term memory, it manages the mail lifecycle through inbox/keep/archive folders, effectively isolating ephemeral coordination information from core knowledge and avoiding token waste.
Limitations
Only supports the local filesystem, with no cloud sync or remote delivery capabilities; employs a best-effort delivery model, with no guarantee that mail will be delivered or read; depends on a fixed workspace directory structure (`.agent-mailbox`).
FAQ
Will failed mail deliveries be retried?
No. The mailbox uses a best-effort model; failed writes are simply skipped with no retry mechanism.
Can multiple agents read from and write to the same mailbox simultaneously?
Yes, but note that concurrent read/write operations on the filesystem may conflict. It is recommended to use simple file naming (e.g., timestamps) to avoid overwrites.
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-7/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-7/raw/index.md