Spec Kit — Spec-Driven Dev Toolkit from GitHub, 85K Stars
GitHub's open-source spec-driven development toolkit. Turn product specs into executable AI agent workflows via the specify CLI instead of vibe coding.
What it is
Spec Kit is an open-source toolkit from GitHub for spec-driven development. Instead of giving AI agents vague instructions (vibe coding), you write structured product specifications and use the specify CLI to turn them into planned, executable tasks. The tool bridges the gap between product requirements and AI-assisted code generation.
Spec Kit targets engineering teams and individual developers who want more control over AI-generated code. By writing specs first, you get predictable outputs from AI agents rather than hoping the agent interprets your intent correctly.
How it saves time or tokens
Spec-driven development reduces the back-and-forth between you and your AI agent. Instead of iterating on vague prompts, you define what you want upfront in a structured spec. The specify CLI breaks specs into discrete tasks that agents can execute sequentially. This approach minimizes wasted tokens on misunderstood requirements and produces more consistent results.
How to use
- Install the specify CLI:
uv tool install specify-cli --from git+https://github.com/github/spec-kit.git. - Initialize a project with
specify init my-projectto create the spec directory structure. - Write a spec document, run
specify planto generate tasks, and execute them with your AI agent.
Example
# Install the CLI
uv tool install specify-cli --from git+https://github.com/github/spec-kit.git
# Initialize a project
specify init my-project
cd my-project
# Create a spec
specify new feature-login
# Plan tasks from the spec
specify plan feature-login
# Execute with your AI agent
specify run feature-login
Related on TokRepo
- Coding AI Tools — AI-powered development tools and workflows
- Prompt Library — Structured prompts and skills for AI agents
Common pitfalls
- Spec Kit requires well-written specifications to be effective. Vague specs produce vague task plans. Invest time in clear requirement writing.
- The tool works best with agents that support structured task execution. Not all AI coding agents integrate equally well.
- Spec Kit is relatively new. The CLI interface and spec format may change between releases. Pin your version for production workflows.
Frequently Asked Questions
Spec-driven development means writing structured product specifications before asking an AI agent to generate code. Instead of ad-hoc prompting, you define requirements, acceptance criteria, and constraints upfront. The AI agent then works from this structured input.
Spec Kit provides a formalized structure for requirements with a CLI that decomposes specs into executable tasks. Plain prompts are unstructured and ambiguous. Specs define what you want, and the CLI plans how to build it step by step.
Spec Kit is designed to work with any AI coding agent. It generates task plans that can be fed to Claude Code, GitHub Copilot, Cursor, or other agents. The specify CLI handles the planning layer while your agent handles execution.
No. Individual developers benefit from spec-driven development too. Writing a brief spec before coding with an AI agent reduces iteration cycles and produces more predictable results, regardless of team size.
Spec Kit generates task plans, not code. The task plans are then executed by your AI coding agent. This separation means you control the planning while the agent handles implementation.
Citations (3)
- Spec Kit GitHub— Spec Kit is GitHub's open-source toolkit for spec-driven development
- GitHub Blog— Spec-driven development as an alternative to unstructured AI prompting
- Anthropic Documentation— AI-assisted software development best practices
Related on TokRepo
Source & Thanks
Discussion
Related Assets
Google Benchmark — Microbenchmark Library for C++
Google Benchmark is a library for measuring and reporting the performance of C++ code with statistical rigor.
Mockito — The Most Popular Java Mocking Framework
Mockito lets Java developers create clean mock objects and verify interactions with a simple, readable API.
Catch2 — Modern C++ Test Framework with Natural Expressions
Catch2 is a header-only C++ testing framework known for its expressive assertion syntax and zero-config setup.