tmux-agents

What This Skill Does

The tmux-agents skill provides an orchestration layer for managing autonomous coding agents within persistent tmux sessions. It allows developers to offload complex or time-consuming coding tasks to background processes, effectively turning their terminal into a multi-agent development environment. Whether you need the high-performance reasoning of Claude or the cost-free experimentation of local Ollama-based models, this skill handles the lifecycle, monitoring, and communication with these agents.

Installation

To integrate this skill into your environment, run the following command in your terminal: clawhub install openclaw/skills/skills/cuba6112/tmux-agents

Ensure you have tmux and Ollama installed on your system. For local agents, remember to pull your preferred coding model (e.g., ollama pull glm-4.7-flash) and run the one-time configuration launch commands provided in the setup documentation.

Use Cases

This skill is ideal for developers juggling multiple project components simultaneously.

• Parallel Development: Work on a frontend feature while an agent handles unit testing or backend refactoring in the background.
• Cost Optimization: Utilize local GPU resources for long-running experiments or heavy refactoring tasks to save on API token costs.
• Asynchronous Coding: Spawn a task, detach, and return hours later to review the progress, preventing IDE blocking and maintaining flow state.
• Production vs. Exploration: Use Cloud-based models for critical production code reviews and local models for speculative refactoring.

Example Prompts

1. "Use the tmux-agents skill to spawn a new claude session named 'api-cleanup' that refactors the existing user service to use the new authentication middleware."
2. "Check the status of my active tmux-agents sessions and give me a brief summary of which ones are still running."
3. "Attach to the 'experiment' session and send a command to the agent asking it to revert the last three file changes it made."

Tips & Limitations

• Session Management: Always name your sessions descriptively to avoid confusion when running multiple concurrent agents.
• Performance: Keep an eye on system resources; while local agents are free, running multiple concurrent heavy LLM processes can impact Mac performance significantly.
• Persistence: Since agents run in tmux, they will persist even if you close your main terminal window, but be aware that they remain active unless explicitly killed.
• API Management: Cloud agents consume API credits; ensure you have sufficient budget or quotas set up if running them in automated loops.