Skill Detail

Supervise tmux-based coding-agent sessions with agent-manager

Use agent-manager to launch, monitor, prompt, revive, and review multiple Claude Code, Codex, OpenCode, Grok, or Gemini CLI sessions running in persistent tmux sessions.

Developer ToolsMulti-Framework
Developer Tools Multi-Framework Security Reviewed
โญ 154 GitHub stars
INSTALL WITH ANY AGENT
npx skills add agentskillexchange/skills --skill supervise-tmux-based-coding-agent-sessions-with-agent-manager Copy
Works best when you want a reusable capability, not another fragile one-off prompt.
At a glance
Tools required
tmux and one supported agent CLI such as Claude Code, Codex, OpenCode, Grok Build, Gemini CLI, or Pi
Install & setup
brew install yoanwai/tap/agent-manager
Author
YoanWai
Publisher
Individual Maintained Open Source
Last updated
Aug 1, 2026
Quick brief

# Supervise tmux-based coding-agent sessions with agent-manager

How it works

What this skill actually does

Use this skill when an operator is coordinating multiple terminal coding agents and needs one durable control surface for session status, prompts, previews, worktrees, and diff review. agent-manager runs supported coding agents inside a private tmux server, keeps sessions alive after the manager closes, and provides a TUI for starting agents, answering blocked prompts, reviving dead sessions, and sending line-comment review feedback back into an agent pane.

## What the agent/operator does

1. Install agent-manager and ensure tmux is available.
2. Run agent-manager from a development machine.
3. Create sessions for Claude Code, OpenCode, Codex, Grok Build, Gemini CLI, or a custom configured CLI tool.
4. Use the session list and live pane preview to find waiting, running, or stuck agents.
5. Press space to send a prompt to the selected session or spawn a new agent in a selected group.
6. Use ctrl+r to review file diffs and send collected review comments back to the agent.
7. Use revive and worktree options to continue interrupted work without losing the session context.

## When to invoke

Invoke this instead of manually juggling terminal tabs when a user has several coding agents working at once, needs to answer permission prompts without attaching to every tmux pane, wants a quick status dashboard, or wants to review changes from agent sessions before integrating them. It is strongest for local engineering workflows where agents run as terminal CLIs and persistence matters.

## Scope boundary

This is not a general tmux tutorial or a generic agent platform listing. Keep the workflow scoped to local tmux-backed coding-agent supervision: launch sessions, monitor status, send prompts, revive sessions, use worktrees, and review diffs through agent-manager. Do not broaden it into hosted orchestration, autonomous team management, or unrelated terminal multiplexing.

## Requirements

– macOS, Linux, or Windows under WSL2.
– tmux.
– At least one supported coding-agent CLI, such as Claude Code, OpenCode, Codex, Grok Build, or Gemini CLI.
– Repository access for sessions that will inspect or modify code.

## Installation

Install with Homebrew on macOS or Linux using brew install yoanwai/tap/agent-manager, or download a release from the upstream GitHub releases page. Then run agent-manager from a shell.

## Verification notes

Source metadata was verified from the exact GitHub repository and README. As of this intake run, YoanWai/agent-manager had 154 GitHub stars, MIT license metadata, push activity on 2026-08-01 UTC, a same-day latest release, docs at agent-manager.dev, and README evidence for supported coding-agent CLIs, tmux session persistence, live status, grouped project tree, quick prompts, revive, worktrees, and full-file diff review.