Maestro: AI Development Orchestrator
Elite-tier orchestration framework for Claude Code CLI. Supercharges AI development through specialized agents, modular skills, intelligent hooks, and persistent memory systems.
Version: 0.6.0
Author: xenitV1 • X/Twitter
Philosophy: "Why over How. Architecture precedes implementation."
Quick Start
Installation
One-line Install (Skills.sh):
npx skills add xenitV1/claude-code-maestro
Manual Plugin Install:
Maestro is distributed as a Claude Code Plugin. To install it manually:
# 1. Add Maestro as a marketplace
/plugin marketplace add xenitV1/claude-code-maestro
# 2. Install the Maestro plugin
/plugin install maestro@xenitV1-claude-code-maestro
Prerequisites
- Node.js 18+ (required for hooks)
- Claude Code CLI
Usage
Since Maestro is a plugin, its commands are namespaced. Use the format /maestro:command.
# Basic orchestration
/maestro your task description
# With Ralph Wiggum (autonomous iterations)
/maestro fix bugs and improve code. ralph 5 iterations
# Design mode
/maestro design new authentication system
# Plan mode
/maestro plan implement user dashboard
# Use the Grandmaster agent directly
/agent:grandmaster
Architecture
┌─────────────────────────────────────────────────────────────────────┐
│ MAESTRO SYSTEM │
├─────────────────────────────────────────────────────────────────────┤
│ ┌──────────┐ ┌──────────────┐ ┌─────────────────────────┐ │
│ │ /maestro │───▶│ grandmaster │───▶│ SKILLS │ │
│ │ command │ │ agent │ │ (frontend, backend, │ │
│ └──────────┘ └──────────────┘ │ tdd, debug, etc.) │ │
│ │ └─────────────────────────┘ │
│ ▼ │
│ ┌─────────────────────────────────────────────────────────────┐ │
│ │ HOOK SYSTEM │ │
│ │ SessionStart │ PostToolUse │ Stop │ PreCompact │ etc. │ │
│ └─────────────────────────────────────────────────────────────┘ │
│ │ │
│ ▼ │
│ ┌──────────────────────────────────────────────────────────────┐ │
│ │ LONG-TERM MEMORY (LTM) │ │
│ │ (brain.jsonl) │ │
│ └──────────────────────────────────────────────────────────────┘ │
└─────────────────────────────────────────────────────────────────────┘
Project Structure
maestro/
├── .claude-plugin/
│ └── plugin.json # Plugin manifest
├── agents/
│ └── grandmaster.md # Central orchestrator agent
├── commands/
│ └── maestro.md # /maestro slash command
├── hooks/
│ ├── hooks.json # Hook configuration
│ ├── lib/ # Shared JS utilities
│ │ ├── utils.js # Cross-platform utilities
│ │ ├── brain.js # LTM operations
│ │ └── ralph.js # Ralph state management
│ ├── session-start.js # Tech stack detection + LTM injection
│ ├── brain-sync.js # LTM sync (PostToolUse)
│ ├── stop.js # Ralph Wiggum iteration
│ ├── ralph.js # QA enforcement
│ ├── sentinel.js # Change detection
│ └── pre-maestro.js # Skill recommendation
├── skills/
│ ├── clean-code/ # Code quality standards
│ ├── frontend-design/ # Elite UI/UX
│ ├── backend-design/ # API & Database patterns
│ ├── tdd-mastery/ # Test-Driven Development
│ ├── debug-mastery/ # Systematic debugging
│ ├── verification-mastery/# Evidence-based completion
│ ├── brainstorming/ # Design-first methodology
│ ├── planning-mastery/ # Implementation planning
│ ├── git-worktrees/ # Isolated workspaces
│ ├── ralph-wiggum/ # Autonomous QA system
│ ├── browser-extension/ # Extension development
│ └── optimization-mastery/# Performance optimization
├── package.json # Node.js metadata
├── LICENSE # MIT License
└── README.md
Memory Systems
Long-Term Memory (brain.jsonl)
Persistent project context across sessions:
- Tech Stack: Frameworks, dependencies, architecture patterns
- Decisions: Key architectural decisions made
- Goals: Project objectives
- Errors: Known issues and blockers
- Compact History: Session summaries after context compaction
- File Changes: Changelog of edits and creates
Ralph Wiggum: Autonomous QA
Elite QA system with Four Pillars:
