ClaudeHub 🛠️

Supercharge your development workflow with systematic research, planning, and autonomous implementation.

The most critical factor in successful LLM-assisted development is alignment: ensuring you and the AI are on the same page before writing code. ClaudeHub's spec-driven workflow gets you there through collaborative research, brainstorming, and detailed planning before autonomous implementation.

✨ Key Features

📋 Spec Driven Workflow

Proven methodology for transforming messy ideas into production-ready code.

• Research & Explore: Deep codebase understanding before making changes
• Brainstorm: Refine rough concepts into validated designs through collaborative dialogue
• Plan: Generate detailed, reviewable implementation plans
• Implement: Execute with built-in review checkpoints and phase verification
• Evaluate: Multi-dimensional quality assessment of your work

Inspired by CipherPowers, Superpowers, and HumanLayer.

🔄 Ralph Wiggum

Your autonomous coding companion that implements features through a structured engineering workflow.

Ralph executes tasks through a Python orchestrator that spawns Claude Code sessions. Each session runs a specialized agent pipeline to complete one task with proper research, implementation, verification, and review.

Key differentiator: Multi-agent pipeline with parallel execution. The Python orchestrator manages session spawning and progress tracking, while specialized agents handle research, coding, testing, and review for each task.

How it works:

# Quick autonomous planning + execution
/claudehub:ralph:run "add user authentication with JWT"

# Or plan interactively first
/claudehub:ralph:plan-deep docs/prd.md
/claudehub:ralph:run

The Agent Pipeline:

3x researcher (parallel) → coder → verifier + reviewer (parallel)

Each task goes through:

1. Research - Three parallel researchers gather context (LOCATING, ANALYZING, PATTERNS)
2. Code - Coder implements + writes tests
3. Verify - Verifier runs tests, linting, type-checking
4. Review - Reviewer assesses code quality
5. Complete - Mark task done in .ralph/TASKS.md, log to .ralph/PROGRESS.md

File Structure:

All Ralph files live in .ralph/:

• ARCHITECTURE.md - Technical approach
• TASKS.md - Tasks with checkboxes, organized by phase
• VERIFICATION.md - Test commands, environment setup
• PROGRESS.md - Learnings from each iteration

Architecture benefits:

• Parallel execution: 3x researcher + verifier/reviewer run concurrently
• Specialized agents: Each agent has a focused responsibility
• Context isolation: Fresh session per task prevents context rot
• Built-in planning: /claudehub:ralph-plan for quick autonomous planning or /claudehub:ralph:plan-deep for interactive deep planning
• Progress tracking: /claudehub:ralph-status shows real-time progress

Perfect for:

• Complex features with clear requirements (PRDs, design docs)
• Projects needing thorough testing and review
• Teams wanting structured engineering workflow
• Long-running implementations with autonomous progress

Inspired by Geoffrey Huntley's Ralph methodology.

🚀 Quick Start

Installation:

/plugin install claudehub@claudehub

# Optional: Install custom statusline
/claudehub:install-statusline
# claudehub │ main │ [Sonnet 4.5] │ ▓▓▓▓▓░░░░░ 50%

Brainstorm an idea:

/claudehub:brainstorm "Add user authentication with OAuth"

Create a detailed plan:

/claudehub:create-plan brainstorm.md

Run Ralph to implement:

# Quick autonomous approach
/claudehub:ralph:run "add user authentication with OAuth"

# Or manual planning first
/claudehub:ralph:plan-deep docs/prd.md
/claudehub:ralph:run

📦 Additional Plugins

GitHub Plugin for streamlined git workflows: intelligent PR creation, branch management, and commit automation.

/plugin install github@claudehub