spec-kitty/spec-kitty

What it solves

Spec Kitty addresses the lack of structure and traceability in AI-assisted coding. It prevents "vibe coding" where requirements, architectural decisions, and acceptance criteria are lost in chat histories, by moving the source of truth for the development process into the Git repository itself.

How it works

It provides a CLI that implements a spec-driven development workflow: spec -> plan -> tasks -> next -> review -> accept -> merge. The tool manages repository-native artifacts (stored in kitty-specs/) and uses isolated git worktrees (stored in .worktrees/) to allow multiple AI agents or developers to work on different work packages in parallel without causing branch conflicts.

Who it’s for

Developers and teams using AI coding agents (such as Claude Code, Cursor, Windsurf, or GitHub Copilot) who need a governed, repeatable software factory approach rather than ad-hoc prompting.

Highlights

• Repo-native governance: Stores specs, plans, and review states directly in Git.
• Agent isolation: Uses git worktrees to prevent "branch chaos" during parallel AI implementation.
• Work-package lifecycle: Tracks tasks through lanes like planned, in_progress, for_review, and approved.
• Broad agent support: Integrates with a wide variety of AI coding tools via slash commands and CLI hooks.
• Retrospective learning: Automatically generates retrospectives after completed missions to improve future work.
• Local-first: Operates locally by default with optional hosted sync integrations.