Agent Mailbox Light

This Skill addresses the need for temporary, low-noise coordination between multiple AI workspaces without committing information to long-term memory. In practical development, developers often need to pass prompts, warnings, or temporary context between different agents or workspaces—for instance, when one workspace detects a system anomaly and needs to notify another, but wants such messages to be automatically archived after the task to avoid polluting core memory.

Usage is straightforward: simply run the script in the sender workspace. It scans other workspaces based on a configured path pattern and writes mail files to the target workspace's .agent-mailbox/inbox/ directory. The recipient workspace automatically checks its inbox directory upon startup or at task entry, reads the latest few emails, moves useful ones to the keep folder, and archives processed emails to the archive folder. The entire process is based on the local filesystem, requiring no network or central service.

It is ideal for developers or teams that need to coordinate multiple independent AI workspaces (e.g., different projects or agents). This is particularly useful for those already using frameworks like OpenClaw, who need to pass temporary, suggestive information between workspaces while strictly controlling token consumption and avoiding固化 all interactions into long-term memory.

We recommend using this tool uniformly when building multi-agent collaboration baselines or when temporary context needs to be preserved across different work sessions. Note that it primarily provides best-effort delivery, with no delivery guarantees, no push notifications, and is not suitable as a task execution bus or command system. Mail content should be treated as reference suggestions for local agents to evaluate.