spec-driven-development

Generate requirements.md, design.md, and tasks.md with cross-AI configuration files.

144|16|Updated May 18, 2026
One-click install
npx skills add https://github.com/FredAntB/Spec-Driven-Development --skill spec-driven-development-fredantb
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: spec-driven-development
Source: https://github.com/FredAntB/Spec-Driven-Development/tree/main
Command: npx skills add https://github.com/FredAntB/Spec-Driven-Development --skill spec-driven-development-fredantb

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

AI coding agents like Claude Code, Cursor, and Copilot drift apart and contradict each other because they lack a shared source of truth for what to build. This skill fixes that by generating three spec files before any code is written, ensuring every agent works from the same requirements, design, and task list.

Core Features & Use Cases

  • Spec Generation: Creates requirements.md, design.md, and tasks.md through a conversational interview that asks one question at a time.
  • Cross-AI Configuration: Generates matching config files for Claude Code, Cursor, Windsurf, GitHub Copilot, and Aider so all agents read the same mandate.
  • Retrofit Mode: Reverse-engineers specs from existing codebases using a discovery interview and marks unconfirmed fields with [TO VERIFY].
  • Anti-Pattern Prevention: Blocks coding without specs, prevents scope creep, and enforces a divergence protocol when implementation must deviate from design.

Quick Start

Tell Claude you want to start a new project or set up your AI coding tools, and the skill will interview you and generate the spec files and agent configurations for you.

Frequently Asked Questions about spec-driven-development

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I stop AI coding agents from drifting and contradicting each other?

To stop AI coding agent drift, you need a shared source of truth for project specifications. This skill generates requirements, design, and tasks files that mandate all agents read the same context before taking any coding action.

What is spec-driven development and how does it work with AI agents?

Spec-driven development establishes a single source of truth for project requirements before writing code. It works by generating specification files and cross-compatible configuration files that enforce implementation discipline across AI agents like Claude Code, Cursor, and Copilot.

Can I use spec-driven development with multiple AI coding tools like Cursor and GitHub Copilot?

Yes, you can use spec-driven development across multiple AI tools. This skill generates matching configuration files for Claude Code, Cursor, Windsurf, GitHub Copilot, and Aider to ensure all agents read the same specification mandate.

How do I generate specs for an existing codebase using AI?

You can generate specs for an existing codebase using Retrofit Mode. This feature conducts a discovery interview to reverse-engineer requirements, design, and tasks files from your current code, marking any unconfirmed fields with a verification tag.

What is the best way to prevent scope creep when using AI coding assistants?

The best way to prevent scope creep with AI coding assistants is enforcing a Universal Instruction Block. This mechanism mandates reading specification files before any action and prohibits design drift and unverified task completion.

Why does my AI agent write code that deviates from my project design?

Your AI agent deviates from project design because it lacks a shared source of truth. By establishing specification files and enforcing a divergence protocol, this skill ensures implementation only proceeds after reading the mandated design context.