Can I resume an AI session after stopping the worker?

Yes, the 'twf stop' command saves the session state, and 'twf resume' reloads the specific Codex session ID to maintain continuity.

How does this skill prevent log interference between multiple workers?

Each worker is assigned a unique ID and an isolated CODEX_HOME directory, ensuring that session logs and history remain separate even when running multiple instances concurrently.

Does this work with the standard Claude CLI?

Yes, the skill includes an optional installation script for custom Claude commands like /cask and /cpend to bridge the two tools.

Does it support hierarchical task management?

Yes, the 'spawn' command allows you to create child workers linked to a parent, which can be viewed and managed as a tree structure.

What is the benefit of using tmux for Codex workers?

Using tmux allows the AI processes to run in the background (headless), provides a way to manually attach to the session for debugging, and makes the workflow highly portable across different environments.

Codex Tmux Workflow Manager

Name: Codex Tmux Workflow Manager
Author: kunhuang12345

bykunhuang12345

•

생산성 및 워크플로

Manages isolated, long-running Codex CLI workers using tmux sessions and automated log polling for seamless background AI workflows.

The tmux-workflow skill provides a robust toolkit for orchestrating multiple persistent Codex CLI instances from external processes like the Claude CLI. By encapsulating Codex within isolated tmux sessions, it enables developers to inject prompts, poll for JSONL-based replies, and manage complex parent-child worker hierarchies without log cross-talk. It is particularly useful for developers who need to run background AI tasks, persist session states across terminal restarts, or build automated multi-agent pipelines that require precise control over environment variables and session history.

주요 기능

01Automated tmux session lifecycle management including start, stop, resume, and health checks.

02Support for hierarchical worker structures with parent-child linking and recursive management.

03Custom Claude CLI command integration for direct interaction using /cask, /cpend, and /cping.

04Isolated worker environments with independent CODEX_HOME directories to prevent log cross-talk.

05Real-time response extraction by polling Codex JSONL logs using inotify or interval-based methods.

061 GitHub stars

사용 사례

01Maintaining persistent, long-running AI context in the background that survives terminal disconnections.

02Orchestrating parallel AI agents to handle different components of a large-scale refactoring project.

03Building automated shell-based workflows that require programmatic interaction with Codex CLI instances.

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add kunhuang12345/team-skill tmux-workflow

For use in Claude.ai and ChatGPT

Download Skill