Superpowers: agentic skills as an operating methodology
A 245,000-star framework for giving coding agents reusable skills. The signal is that agent capability is becoming a managed artifact, like code.
Superpowers is an agentic skills framework and software development methodology sitting at roughly 245,000 GitHub stars and still adding over a thousand a day. The idea: instead of prompting an agent from scratch each time, you give it a library of versioned, reusable skills that encode how work should be done in your environment.
The enterprise signal here is bigger than the repo. Skills are agent capability expressed as a managed artifact: written down, versioned, reviewed, and improved over time. That is the same shift that turned infrastructure into code a decade ago, and it lands directly in the engineering factory thesis. When your agents’ working methods live in files rather than in individual developers’ prompting habits, you can standardize them across teams, audit them, and treat improvements to the factory as commits rather than tribal knowledge.
Popularity is not validation, and this is the caveat: a methodology framework this popular becomes its own kind of lock-in, and a 245,000-star general-purpose skill library was not written with your architecture, your compliance rules, or your approval gates in mind. Adopt the pattern before the package. The first step is to write three or four skills that encode your organization’s own non-negotiables (migration policy, test evidence, review requirements) and measure whether agents following them produce fewer correction cycles. If they do, you have learned the real lesson of this repo, whether or not you ever install it.
Written by Adib Kadir. Product and engineering executive focused on rolling out AI at enterprise scale.
Start a conversation