Priivacy-ai/spec-kitty

What it solves

Spec Kitty provides a structured, spec-driven development workflow for AI coding agents, preventing the loss of requirements, architecture decisions, and acceptance criteria that often occur during ad hoc "vibe coding" sessions. It transforms product intent into a traceable, governed process where the repository remains the source of truth for all mission state.

How it works

It operates as a CLI tool that implements a lifecycle for AI coding: spec -> plan -> tasks -> next -> review -> accept -> merge. The tool creates repository-native artifacts (under kitty-specs/) to track specs, plans, and work packages. To prevent branch chaos when using multiple agents in parallel, it utilizes isolated git worktrees (under .worktrees/). It also provides a governance layer via the dispatch command to map operator intent to runtime behavior and a local kanban dashboard for progress tracking.

Who it’s for

Developers and teams building "software factories" who use AI coding agents (such as Claude Code, Cursor, Gemini, or GitHub Copilot) and want a repeatable, governed workflow with clear boundaries and audit trails rather than a black-box autonomous process.

Highlights

• Repo-native state: Stores specs, plans, and reviews directly in Git, ensuring the repository is the source of truth.
• Git worktree isolation: Allows multiple agents to work in parallel without interfering with each other's branches.
• Lifecycle lanes: Manages work packages through specific states: planned, in_progress, for_review, approved, and done.
• Broad agent support: Integrates with a wide variety of AI coding tools including Claude Code, Codex, Cursor, Gemini, Copilot, Windsurf, and more.
• Retrospective learning: Automatically generates retrospectives for completed missions to synthesize improvements for future work.