GitHub Spec Kit: Write Specs First, Then Code

GitHub open-sourced Spec Kit, a toolkit for spec-driven development that has over 110,000 stars. The idea is simple: instead of vibe coding and hoping AI guesses your intent, you write a clear spec first, then AI implements according to that spec.

What it is

Spec Kit is an open-source toolkit maintained by GitHub that helps you practice Spec-Driven Development. Instead of telling an AI agent to "code this feature" and hoping it guesses the main flow, data model, error handling, and definition of done correctly, you write a structured spec that defines what to build, why it matters, what the constraints are, and what the acceptance criteria are. The spec then becomes executable, meaning it directly generates implementation code rather than just sitting in a document.

The toolkit includes a CLI called specify, installed via uv, and a seven-step workflow: constitution, specify, plan, tasks, implement, and converge. Each step corresponds to a slash command that you run inside your AI coding agent. Spec Kit is compatible with over 30 AI coding agents, including Claude Code, GitHub Copilot, Codex CLI, Cursor, Windsurf, Cline, and Aider, so you are not locked into one tool.

Why it matters

• It separates the thinking phase from the coding phase, which reduces the amount of guessing the AI has to do.
• The structured workflow, from constitution to converge, gives you a repeatable process instead of ad-hoc prompting.
• Compatibility with over 30 AI coding agents means no tool lock-in. You can switch agents without changing your spec workflow.
• It is an official GitHub repo with over 110,000 stars, not an anonymous side project. That gives it community validation and a real maintenance commitment.
• The bundle system packages extensions and presets for each role: product manager, developer, security researcher, and business analyst. One specify bundle install command sets up everything for your role.

How it works

The workflow has seven steps. Constitution defines project principles. Specify describes what to build. Plan chooses the tech stack. Tasks breaks the work into a task list. Implement has the AI code according to the plan. Converge checks whether the code matches the spec. Each step is a slash command: /speckit.constitution, /speckit.specify, /speckit.plan, /speckit.tasks, /speckit.implement. You run these inside your AI coding agent, and Spec Kit generates the appropriate context and instructions for the agent to follow.

To get started, you install the CLI with uv tool install specify-cli, initialize a project with specify init, choose your integration, and the slash commands appear in the agent you are using.