Spec Kit is an open source toolkit that gives AI coding agents structured
processes, reusable templates, and documented outcomes. Start with a ready-made
process, customize it, or bring your own.
A process adapted or assembled from reusable primitives
These are independent entry points, not mandatory phases. SDD ships in core;
bug fixing and assessment are bundled extensions you install when needed, and
the primitives below let you adopt or create other processes.
Capabilities
Spec Kit combines ready-to-use processes with composable mechanisms:
Search the active catalogs with commands such as specify extension search,
specify preset search, specify workflow search, and specify bundle search.
Spec Kit includes community catalogs, and organizations can add their own
curated sources. Catalog listings are not audits or endorsements, so review
community source code before installing or running it.
Get started
You need Python 3.11+, uv,
and a supported AI coding agent on Linux, macOS, or Windows.
For CLI setup only, run this in your terminal to install Spec Kit and create a project:
The examples use GitHub Copilot's default skills mode. Replace copilot with
your integration key
to use another supported agent.
Already have code? Follow the
existing-project guide.
For pinned releases, other installers, CI, or troubleshooting, see
Installation.
To update an existing installation, see Upgrade.
Now launch your coding agent in the project directory and choose a process
below. Invoke each /speckit-*skill in your agent's chat, one at a time,
and review the result before continuing. These are agent skills, not terminal
commands. Other agents and modes may use
different invocation syntax.
Spec-Driven Development
Define what and why before deciding how to build it. SDD turns your
requirements into a specification, a technical plan, and actionable tasks,
then guides implementation against those artifacts.
Constitution once per project; specify → plan → tasks → implement → converge per feature.
Invoke these skills in your agent's chat:
/speckit-constitution Create principles focused on code quality, testing, and maintainability.
/speckit-specify Build a photo organizer with albums grouped by date and a tile preview of each album.
/speckit-plan Use Vite with vanilla JavaScript. Keep images local and store metadata in SQLite.
/speckit-tasks
/speckit-implement
/speckit-converge
Repeat implement → converge until convergence reports Converged.
Add clarification, checklists, and consistency analysis when you need extra
quality gates.
Keep diagnosis, repair, and verification separate so the agent fixes the assessed
cause and checks the original symptom. No SDD feature workflow is required first.
CLI setup (terminal): install the opt-in extension from the project directory:
specify extension add bug
Then invoke the assess → fix → test skills in your agent's chat:
/speckit-bug-assess "Submitting an empty password crashes the login form." slug=login-crash
/speckit-bug-fix slug=login-crash
/speckit-bug-test slug=login-crash
The reports live in .specify/bugs/login-crash/. Review the final verdict:
verified, partial, or failed. Missing verification is not a successful fix.
Gather evidence before committing to an idea, whether or not it becomes software.
This standalone process works even in a project with no source code.
CLI setup (terminal): install the opt-in extension from the project directory:
specify extension add assess
Then invoke the intake → research → define → shape → decide skills in your agent's chat:
/speckit-assess-intake "Let users work offline and sync when they reconnect." slug=offline-mode
/speckit-assess-research slug=offline-mode
/speckit-assess-define slug=offline-mode
/speckit-assess-shape slug=offline-mode
/speckit-assess-decide slug=offline-mode
The artifacts live in .specify/assessments/offline-mode/, ending in a
go / needs-clarification / kill decision. Resolve unknowns by refining the
existing Markdown artifacts directly or with the agent, rather than regenerating
whole stages. A go decision can be handed to /speckit-specify if you choose
to build it; stopping with a documented reason is also a useful result.
