Spec Kit gives AI coding agents structured processes, reusable templates, and documented outcomes. A command-line tool sets up a project for the chosen agent, and the processes then run as skills you invoke one at a time in the agent's chat, reviewing each result before continuing.
| What you need | Process | Outcome |
|---|---|---|
| Build a feature or application | Spec-Driven Development (SDD) | A specification carried through planning, implementation, and convergence |
| Diagnose and repair broken behavior | Bug fixing | An assessed cause, scoped fix, and recorded verification |
| Decide whether an idea deserves investment | Idea assessment | An evidence-backed go, clarify, or stop decision |
The three processes are independent entry points, not mandatory phases. SDD ships in core: a project constitution is written once, then each feature moves through specify, plan, tasks, implement, and converge, with implement and converge repeated until convergence reports the feature as converged. Each phase produces a Markdown artifact that feeds the next, so the agent works from structured context instead of ad hoc prompts. Bug fixing and idea assessment are bundled, opt-in extensions, and assessment works even in a project with no source code.
The process itself lives in extensions, presets, workflows, and bundles, so teams can tune SDD, replace it, or run processes outside software development. Community processes cover areas such as product management, .NET Framework migration, multi-agent orchestration with quality gates, and long-form fiction writing.
Features
- Spec-Driven Development: constitution, specify, plan, tasks, implement, and converge skills, with optional clarification, checklists, and consistency analysis
- Bug fixing extension: assess, fix, and test skills that keep diagnosis, repair, and verification separate and end in a verified, partial, or failed verdict
- Idea assessment extension: intake, research, define, shape, and decide skills ending in a go, needs-clarification, or kill decision
- Agent integrations: 38 coding agents, including Copilot, Gemini, Codex, Kilo Code, Zed, Claude, and Kiro, plus a generic integration for any other tool
- Customization: extensions add capabilities, presets adapt behavior, workflows automate steps, and bundles package a role-based setup
- Community catalog: community extensions and presets, including alternative processes such as AIDE, Canon, Product Forge, and MAQA
- Self-hosted catalogs: works offline and behind firewalls, and organizations can host their own catalogs of integrations, extensions, presets, workflows, and bundles
