Vibe-Switch
tmux for AI Agents — Orchestrate multiple AI coding agents in parallel with one command. Run Claude Code, Codex CLI, and Gemini CLI simultaneously on isolated Git branches with seamless context handoff.
Install via CLI (Recommended)
clawhub install openclaw/skills/skills/brianjhang/vibe-switchVibe-Switch — tmux for AI Agents
Orchestrate multiple AI coding agents in parallel with one command. Each agent runs in its own isolated Git worktree and branch. When one agent finishes, hand off the context seamlessly to another.
Install
npm install -g vibe-switch
Quick Start
# Start agents in parallel
vibe run "Implement JWT auth middleware" --agent gemini
vibe run "Build responsive login page" --agent codex
# Monitor all agents
vibe status
vibe watch
# Hand off context between agents
vibe handoff vibe/gemini-c3d5 --to codex -m "API is ready at /api/auth"
# Cleanup
vibe stop --all
vibe clean
Command Reference
| Command | Description |
|---|---|
vibe run "<task>" --agent <agent> | Start an agent on an isolated Git branch + worktree |
vibe status | Display all tasks, status, and branches |
vibe watch | Real-time multi-agent output streaming |
vibe log <branch> [-f] | View or follow logs for a specific task |
vibe stop [branch|--all] | Stop one or all agents |
vibe handoff <branch> --to <agent> | Transfer context to another agent |
vibe summary <branch> | View task summary, logs, and diff stat |
vibe clean | Clean up completed tasks, logs, and worktrees |
vibe agents | List installed agent adapters |
vibe doctor | Diagnose environment and agent availability |
vibe init | Create project config .vibeswitch.json |
vibe config [key] [value] | View or set project config |
Supported Agents
| Agent | Command | Sandbox | Best For |
|---|---|---|---|
| Claude Code | claude | No | Architecture, code review, broad reasoning |
| Codex CLI | codex | Yes | Implementation, refactors, test-driven changes |
| Gemini CLI | gemini | No | Exploration, cross-checking, documentation |
| Antigravity | antigravity | No | Full-stack development, complex tasks |
| OpenClaw | openclaw | No | General purpose AI coding |
Architecture
- Git Worktree Isolation: Each agent runs in a dedicated branch and directory — no file conflicts.
- Adapter Pattern: Modular design for easy integration of new AI agent CLIs.
- JSON File Storage: Task metadata persisted locally at
~/.vibe-switch/— no database needed.
Common Workflow Patterns
Parallel Development
vibe run "Implement auth API and migration" --agent claude --branch vibe/auth-api
vibe run "Build auth UI and form validation" --agent codex --branch vibe/auth-ui
vibe run "Write integration tests for login" --agent gemini --branch vibe/auth-tests
vibe status
vibe watch
Context Handoff
# Gemini finished the API → hand off to Codex for the frontend
vibe handoff vibe/auth-api --to codex -m "API complete. Build the client-side integration."
Agent Capability Matrix
Metadata
Not sure this is the right skill?
Describe what you want to build — we'll match you to the best skill from 16,000+ options.
Find the right skillPaste this into your clawhub.json to enable this plugin.
{
"plugins": {
"official-brianjhang-vibe-switch": {
"enabled": true,
"auto_update": true
}
}
}Tags
Related Skills
designer-intelligence-station
Designer intelligence collection tool. Monitors 46 public sources (AI/hardware/mobile/design), dynamic quality-based filtering v2.1.8, generates structured daily/weekly reports. All data stored locally.
Agent Migration Pack
将AI Agent完整迁移到新环境或分享给其他用户的标准化工具包，包含身份、记忆、技能、风格等完整信息，支持状态迁移
Claude Code CLI for OpenClaw
Install, authenticate, and use Claude Code CLI as a native coding tool for any OpenClaw agent system.
Reflection Engine
Analyzes an AI agent's dream/knowledge graph to surface the top recurring themes from its inner life. Reads DreamInsight entities from a knowledge_graph.json and returns the most frequent conceptual patterns — giving your agent a window into what it's actually thinking about. Built and proven on Albion, an autonomous AI running 31,000+ dream cycles on a Raspberry Pi 5.
project-evaluator
描述一个项目想法，AI 从市场/技术/商业/风险四个维度系统评估， 输出评估报告、竞品速查、MVP建议，帮你决策「值不值得做」。