What it is: Spec Kit is GitHub’s open-source toolkit for spec-driven AI development. Four phases (Spec → Plan → Tasks → Implement). Acceptance criteria become machine-verifiable quality gates. Supports Claude Code, Copilot, Cursor, Codex, Gemini CLI, Tabnine, and ~30 other coding agents.
Who it is for: Engineering teams using AI coding agents on real codebases who want structured, predictable output.
Best if: You want a short reference on what Spec Kit does and how it fits with AGENTS.md / feature lists.
Skip if: You only use AI for ad-hoc, one-off edits. Want one practical AI workflow every morning? Subscribe to our free daily newsletter.
What is Spec Kit?
Spec Kit is GitHub’s open-source toolkit for spec-driven AI development. Its core idea: turn fuzzy product requirements into structured artifacts (Spec → Plan → Tasks → Implement), each acting as a quality gate the AI coding agent must satisfy before moving to the next phase. The /speckit.tasks command generates a tasks.md with dependency ordering, parallel-execution markers, exact file paths, and checkpoint validation between phases.
Why does Spec Kit matter?
Spec Kit codifies the harness-engineering pattern that has become industry standard in 2026: acceptance criteria locked upfront, the agent works against the spec, the spec is verifiable. It supports more agents than any other comparable toolkit — Claude Code, GitHub Copilot, Cursor CLI, Gemini CLI, Codex CLI, Tabnine, and roughly 30 others. For teams, it’s the closest thing to a standard process for AI-assisted development.
How does Spec Kit work?
Four phases. Spec: describe what the project should do, in plain language. Plan: translate the spec into an implementation strategy. Tasks: break the plan into ordered tasks with exact file paths, dependency markers, and parallel-execution flags. Implement: the agent executes the tasks one at a time, validating against the spec at each checkpoint.
Functionally, the tasks artifact is Spec Kit’s equivalent of feature_list.json — locked acceptance criteria the agent works against. The implement phase can’t reach beyond the tasks file. See our Feature Lists for AI Coding Agents guide for the full pattern.
Related terms
Learn more on Beginners in AI
Sources and further reading
Last reviewed: May 2026. AI terminology evolves quickly — verify specifics on the official source pages above.
Get Smarter About AI Every Morning
Free daily newsletter — one term, one tool, one tip. Plain English.
Free forever. Unsubscribe anytime.
You may also like
- Feature Lists for AI Coding Agents
- Harness Engineering for Beginners
- Claude Code Best Practices
- Feature list
- AGENTS.md
- Harness engineering
- AI Glossary
Two ways to go further
The AI Prompt Library
1,000+ ready-to-use prompts for Claude, ChatGPT, and Gemini. Stop staring at a blank box.
Get it for $39 →2-Hour Live AI Crash Course
A private, beginner-friendly session across Claude, ChatGPT, Gemini, and the wider landscape.
Book for $125 →