spec-kit is at 128k stars and shipping a release almost every day
GitHub's spec-kit picked up 1,147 stars in a day on its way past 128,000, which is remarkable for a repo whose entire proposition is that you should write the spec first. The pitch is Spec-Driven Development: define what to build before building it, with any AI coding agent. MIT licensed, Python, 1,795 commits, and it supports 30-plus coding agents rather than betting on one.
The workflow is a chain of commands that each produce an artifact. Constitution, then specification, then technical plan, then task breakdown, then implementation. The core belief is that single-prompt code generation is the wrong unit of work and multi-step refinement with a written contract at each stage is the right one. If you have watched an agent confidently build the wrong feature for forty minutes, you already agree.
What is more interesting than the concept is the release cadence. Look at the log for this month and it is 0.15.0, 0.15.1, 0.15.2, 0.16.0, 0.16.1, 0.16.2, 0.16.3, 0.16.4 β a release every day or two. And the content of those releases says where the project's center of gravity moved. Copilot integration switched its default from commands to skills. First-class agent-native runtime hooks landed. Extension manifests can now ship their own templates and scripts. Archive downloads for extensions and presets got security hardening. There is a whole preset ecosystem now covering security, architecture, iSAQB, accessibility, cross-platform and agent-parity governance, with presets resolvable in order and namespaced as self-contained commands.
Read that list again. It is not a prompt template collection anymore, it is a plugin runtime with a governance layer and a supply-chain security problem it is actively patching. The trajectory from "here is a nice way to structure your prompts" to "here is a package manager for agent behavior with signed archives" took about four months, and it is the same trajectory DeepSeek's harness and every cockpit tool is on.
github.com/github/spec-kit, docs at github.github.com/spec-kit.
← Back to all articles
The workflow is a chain of commands that each produce an artifact. Constitution, then specification, then technical plan, then task breakdown, then implementation. The core belief is that single-prompt code generation is the wrong unit of work and multi-step refinement with a written contract at each stage is the right one. If you have watched an agent confidently build the wrong feature for forty minutes, you already agree.
What is more interesting than the concept is the release cadence. Look at the log for this month and it is 0.15.0, 0.15.1, 0.15.2, 0.16.0, 0.16.1, 0.16.2, 0.16.3, 0.16.4 β a release every day or two. And the content of those releases says where the project's center of gravity moved. Copilot integration switched its default from commands to skills. First-class agent-native runtime hooks landed. Extension manifests can now ship their own templates and scripts. Archive downloads for extensions and presets got security hardening. There is a whole preset ecosystem now covering security, architecture, iSAQB, accessibility, cross-platform and agent-parity governance, with presets resolvable in order and namespaced as self-contained commands.
Read that list again. It is not a prompt template collection anymore, it is a plugin runtime with a governance layer and a supply-chain security problem it is actively patching. The trajectory from "here is a nice way to structure your prompts" to "here is a package manager for agent behavior with signed archives" took about four months, and it is the same trajectory DeepSeek's harness and every cockpit tool is on.
github.com/github/spec-kit, docs at github.github.com/spec-kit.
Comments