Spec-Driven Development toolkit for specification-first workflows with AI agents. Generate executable specifications, implementation plans, and task breakdowns that drive code generation. Use when building software projects requiring structured specification workflows, integrating AI coding assistants with spec-driven methodologies, or migrating to specification-first practices.
Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.
Une commande directe contourne le prompt de vérification. Examinez la source avant de l'exécuter.
Spec-Driven Development toolkit for specification-first workflows with AI agents. Generate executable specifications, implementation plans, and task breakdowns that drive code generation. Use when building software projects requiring structured specification workflows, integrating AI coding assistants with spec-driven methodologies, or migrating to specification-first practices.
Spec Kit is an open-source toolkit by GitHub that enables Spec-Driven Development (SDD) — a methodology where specifications become the primary artifact and code serves as their generated expression. It provides the specify CLI to bootstrap projects, plus slash commands (/speckit.*) that integrate with 30+ AI coding agents to automate the specification → planning → tasking → implementation workflow.
The core idea: instead of writing code directly from prompts ("vibe coding"), you first create rich specifications, then derive technical plans and executable tasks from them. This eliminates the gap between intent and implementation by making specifications precise enough to generate working systems.
When to Use
Starting a new project with structured specification-first development
Integrating AI coding assistants (Claude Code, Copilot, Gemini CLI, Codex, Pi, etc.) with SDD workflows
Creating governing principles (constitution) for a development project
Breaking complex features into specifications, plans, and actionable tasks
Extending Spec Kit with custom commands via the extension system
Customizing Spec Kit behavior via presets (template/command overrides)
Migrating existing projects to specification-driven practices
Publishing extensions or presets to the community catalog
Core Concepts
Spec-Driven Development (SDD)
SDD inverts the traditional power structure: specifications don't serve code — code serves specifications. The Product Requirements Document (PRD) is the source that generates implementation. Technical plans are precise definitions that produce code. This eliminates the gap between intent and implementation.
Key principles:
Intent-driven development — team intent expressed in natural language, design assets, and core principles
Executable specifications — precise, complete, unambiguous enough to generate working systems
Continuous refinement — consistency validation as an ongoing process, not a one-time gate
Research-driven context — agents gather technical context throughout specification
Bidirectional feedback — production reality informs specification evolution
Branching for exploration — multiple implementation approaches from the same specification
The Six-Step Workflow
specify init — Bootstrap the project with Spec Kit structure and agent integration
/speckit.constitution — Define governing principles and development guidelines
/speckit.specify — Create feature specifications (what and why, not how)
/speckit.plan — Generate technical implementation plans with architecture choices
/speckit.tasks — Break down into actionable, parallelizable task lists
/speckit.implement — Execute all tasks to build the feature
Optional quality gates:
/speckit.clarify — Resolve ambiguities in specifications before planning
/speckit.analyze — Cross-artifact consistency and coverage analysis
Spec Kit uses automatic feature numbering (001, 002, 003…) and creates semantic git branches for each feature. The active feature is detected from the current branch name. Specs live in specs/<feature-number>-<name>/ directories.
Constitution
The constitution (memory/constitution.md) is a set of immutable architectural principles governing how specifications become code. It includes articles covering library-first design, CLI interface mandates, test-first development, simplicity gates, anti-abstraction rules, and integration-first testing.
Extensions and Presets
Extensions add new capabilities (new commands, integrations, workflows)
Spec Kit is installed from GitHub (not PyPI). The official package is specify-cli.
Persistent Installation (Recommended)
# Pin a specific release tag for stability
uv tool install specify-cli --from git+https://github.com/github/spec-kit.git@vX.Y.Z
# Or with pipx
pipx install git+https://github.com/github/spec-kit.git@vX.Y.Z
# New project directory
specify init <PROJECT_NAME>
# Current directory
specify init . --integration copilot
# Force merge into non-empty directory
specify init . --force --integration claude
# Skip git initialization
specify init my-project --no-git
# Specify script type (sh or ps)
specify init my-project --script sh
Prerequisites
Linux, macOS, or Windows (PowerShell scripts supported)
Python 3.11+
Git
uv (recommended) or pipx for package management
Supported AI coding agent
Usage Examples
Complete Workflow: Building a Photo Album App
# Step 1: Initialize project
specify init photo-albums --integration claude
# Step 2: Define constitution (in your coding agent)
/speckit.constitution Library-first approach, strict TDD, functional programming patterns
# Step 3: Create specification
/speckit.specify Build an application to organize photos in albums grouped by date. Albums are flat (no nesting). Photos previewed in a tile interface with drag-and-drop reorganization.
# Step 4: Clarify ambiguities
/speckit.clarify Focus on security and performance requirements
# Step 5: Create technical plan
/speckit.plan Vite with vanilla HTML/CSS/JS. SQLite for metadata. No image uploads — local storage only.
# Step 6: Generate tasks
/speckit.tasks
# Step 7: Analyze (optional quality gate)
/speckit.analyze
# Step 8: Implement
/speckit.implement
Using Extensions
# Search available extensions
specify extension search jira
# Install from catalog
specify extension add spec-kit-jira
# List installed extensions
specify extension list
# Remove an extension
specify extension remove spec-kit-jira --keep-config
