agent-teams

This Skill addresses the challenges of orchestrating and facilitating communication among multiple Claude processes during collaborative work. In practical development, when multiple AI agents are needed to handle different modules in parallel (such as frontend/backend separation or multi-file code review), developers often lack a lightweight coordination mechanism, while using databases or network layers proves too cumbersome.

The usage is straightforward: simply execute the claude command in your terminal with a team identifier parameter to create independent agent processes. Each agent runs in a separate tmux pane, exchanging messages via JSON files, with file locks (fcntl) ensuring concurrency safety. You can then assign tasks using natural language instructions, such as "Have the frontend agent implement the login page and the backend agent implement the login API." The system automatically creates team configurations, launches members, and monitors the task queue.

It is particularly well-suited for teams or individuals engaged in parallel development and multi-agent collaboration, especially those already familiar with tmux and the Claude CLI. For large-scale refactoring, code reviews, or cross-file implementations that involve handling multiple loosely-coupled subtasks simultaneously, it significantly boosts efficiency and avoids the context window limitations of a single agent.

We recommend using this tool when tasks have inherent parallelism, such as independent multi-module development or competitive analysis. Note that it is primarily designed for loosely-coupled tasks and is not suitable for scenarios requiring tight, sequential reasoning. Additionally, it requires the host environment to have tmux and the Claude CLI installed, with experimental features enabled.