RefSpec Kit - GitHub's Spec-Driven Development Toolkit

Spec Kit - GitHub's Spec-Driven Development Toolkit

Spec Kit - GitHub's Spec-Driven Development Toolkit

Build high-quality software faster. Focus on product scenarios and predictable outcomes instead of vibe coding every piece from scratch.

What Is It?

Spec Kit is GitHub's open source toolkit for Spec-Driven Development (SDD). It's a structured process where specifications become executable, directly generating working implementations rather than just guiding them.

Core Philosophy: Specifications define the "what" before the "how" through multi-step refinement rather than one-shot code generation from prompts.

What Problem Does It Solve?

Traditional Development Issues

  • Vibe coding from scratch without clear requirements
  • Lost context in one-off prompts
  • Code-first approach where specs are just scaffolding
  • Disconnect between what you want and what you get

Spec-Driven Development Solution

  • Intent-driven: Specifications define requirements before implementation
  • Rich specs: Using guardrails and organizational principles
  • Multi-step refinement: Systematic process from spec → plan → tasks → implementation
  • Predictable outcomes: Clear validation and acceptance criteria

How to Use It Quickly

Quick Start

# Install (recommended - persistent)
uv tool install specify-cli --from git+https://github.com/github/spec-kit.git

# Create new project
specify init <PROJECT_NAME>

# Or initialize in existing project
specify init . --ai claude
# or
specify init --here --ai claude

# Check installed tools
specify check

The 6-Step Workflow

Once initialized, your AI agent gets /speckit.* slash commands:

1. Establish Principles (/speckit.constitution)

/speckit.constitution Create principles focused on code quality, 
testing standards, user experience consistency, and performance requirements

2. Create Spec (/speckit.specify)

/speckit.specify Build an application that can help me organize my photos 
in separate photo albums. Albums are grouped by date and can be re-organized 
by dragging and dropping on the main page.

3. Clarify (/speckit.clarify - optional but recommended)

  • Structured questioning to fill gaps before planning

4. Create Plan (/speckit.plan)

/speckit.plan The application uses Vite with minimal number of libraries. 
Use vanilla HTML, CSS, and JavaScript. Images stored in local SQLite database.

5. Break Down Tasks (/speckit.tasks)

/speckit.tasks

6. Implement (/speckit.implement)

/speckit.implement

Supported AI Agents (20+)

  • ✅ Claude Code
  • ✅ Cursor
  • ✅ Codex CLI
  • ✅ GitHub Copilot
  • ✅ Windsurf
  • ✅ Gemini CLI
  • ✅ opencode
  • ✅ Qoder CLI
  • ✅ Auggie CLI
  • ✅ Roo Code
  • ✅ Amp
  • ✅ IBM Bob
  • ✅ Jules (by Google)
  • ✅ Kilo Code
  • ✅ SHAI (OVHcloud)
  • ⚠️ Amazon Q Developer CLI (limited - no custom args for slash commands)
  • …and more

Development Phases

0-to-1 Development (Greenfield)

  • Start with high-level requirements
  • Generate specifications
  • Plan implementation steps
  • Build production-ready applications

Creative Exploration

  • Explore diverse solutions
  • Support multiple tech stacks & architectures
  • Experiment with UX patterns

Iterative Enhancement (Brownfield)

  • Add features iteratively
  • Modernize legacy systems
  • Adapt processes

Key Features

Constitution-First

Project governing principles guide all decisions throughout development (stored in .specify/memory/constitution.md)

Multi-Artifact Structure

.specify/
├── memory/
│   └── constitution.md
├── scripts/
│   ├── create-new-feature.sh
│   ├── setup-plan.sh
│   └── update-claude-md.sh
├── specs/
│   └── 001-feature-name/
│       ├── spec.md
│       ├── plan.md
│       ├── tasks.md
│       ├── data-model.md
│       ├── contracts/
│       └── research.md
└── templates/
    ├── spec-template.md
    ├── plan-template.md
    └── tasks-template.md

Branch-per-Feature

  • Git branches created per feature (e.g., 001-create-taskify)
  • Enables parallel exploration of different implementations
  • PR-ready workflow with GitHub CLI integration

Validation & Checklists

  • Review & Acceptance Checklist in spec
  • Structured validation before implementation
  • Cross-artifact consistency analysis (/speckit.analyze)

Prerequisites

  • Linux/macOS/Windows
  • uv for package management
  • Python 3.11+
  • Git
  • Supported AI coding agent

Research Goals

GitHub is using Spec Kit to validate:

  1. Technology independence - Works across diverse tech stacks, languages, frameworks
  2. Enterprise constraints - Mission-critical apps with organizational requirements
  3. User-centric development - Different cohorts, from vibe-coding to AI-native
  4. Creative & iterative processes - Parallel exploration, upgrades, modernization

Comparison to Other Tools

See moc-planning-work for how Spec Kit compares to openspec.dev and other planning tools.

By

GitHub - Den Delimarsky (@localden) & John Lam (@jflam)

MIT License

Edit this page