The SpecKit workflow follows GitHub’s Spec Kit method and writes its files in Spec Kit’s folder layout, so tools and teams that already use Spec Kit can pick them up. In the picker it is SpecKit generate-greenfield (v1).
Unlike MySpec, SpecKit keeps one constitution for the whole project and writes a numbered feature folder for every session.
At a glance
- Best for: New features, when your team or tools follow Spec Kit
- Project type: Greenfield (or Auto)
- Stages: Interview → Constitution → Specify → Plan → Tasks
- Files:
specs/memory/constitution.md, shared by the project;spec.md,plan.mdandtasks.mdinspecs/<number>-<feature>/
The first SpecKit session in a project writes specs/001-<feature>/, the next
writes specs/002-<feature>/, and so on.
Stages
Interview
The same product questions as in MySpec: the problem, the users, the core features and journeys, and the scope of the first version.
Constitution
The first SpecKit session in a project writes specs/memory/constitution.md:
numbered core principles (a principle can be marked non-negotiable), additional
constraints, the development workflow and how the constitution is governed.
A later SpecKit session in the same project doesn’t start a new constitution. The AI treats the existing one as the rules, skips its questions and updates it if the new feature needs a change. You review the update as usual.
Specify
Collecting asks for one to three prioritized user stories (P1 is the smallest useful slice, your MVP), how each can be tested on its own, what the system must do for each story, measurable success criteria, key entities, edge cases and assumptions.
Generating writes spec.md: what to build and why, without implementation
details. It has user scenarios with priorities and acceptance scenarios,
functional requirements (FR-001 …), success criteria (SC-001 …), key
entities, edge cases and assumptions.
Review checks that the spec is complete and follows the constitution, and it won’t pass while any question is marked as still needing clarification.
Plan
Collecting asks for the technical context the spec and constitution don’t pin down: language and version, main frameworks, storage, testing tools, target platform, project type, and performance goals and constraints.
Generating writes plan.md: a summary, the technical context, a
Constitution Check, the architecture with a diagram, system modules, the data
model, a testing strategy and key decisions. Optional sections cover the API,
security, deployment and observability when the feature needs them.
Review re-checks the plan against every constitution principle. Each conflict must be resolved, or justified under Complexity Tracking, before the plan passes.
Tasks
Collecting is short: whether to include test tasks (Spec Kit treats tests as optional), whether the project uses the default layout, and any extra phases you want.
Generating writes tasks.md in phases: setup, foundational work, one phase
per user story in priority order (the first is the MVP), and a final polish
phase. Each task has an ID such as T001; [P] marks tasks that can run in
parallel, and a story tag such as [US1] links a task to its user story.
Review checks that every requirement, success criterion and user story is covered, and that the tasks follow Spec Kit’s conventions.
Tips
- Run one feature per session. Each session adds a feature folder and reuses the project’s constitution, which is how Spec Kit is meant to grow.
- Keep the constitution stable. Every later SpecKit session works within it. If you change it, later features follow the new rules.
- Going back.
/gobacklists the stages you can return to. You can’t go back to the Interview.
When to pick another workflow
If nobody on your team needs the Spec Kit layout, MySpec writes a single bundle per session with a fuller requirements file and a solution file.
See Spec files for what each file contains.