Claude Superpowers

Most coding agents start typing the moment you ask. This one makes yours stop, agree a spec with you, plan the work, then build it test-first.

Superpowers is a free, MIT-licensed framework that gives coding agents such as Claude Code a complete development methodology. Rather than writing code immediately, the agent works a specification out of the conversation, turns it into an implementation plan you approve, then builds it test-first using subagents. In Claude Code you install it by running /plugin install superpowers@claude-plugins-official, and the skills then trigger on their own.

What you get

  • A whole methodology rather than a prompt pack: spec, then plan, then build
  • Skills that fire automatically, so there are no new commands to memorise
  • Red/green TDD, YAGNI and DRY enforced by the process instead of by nagging
  • Install steps for Claude Code, Codex, Cursor, Gemini CLI, Copilot CLI and more
  • MIT licensed, so you can read every skill and change any of them

Who this is not for

This is not a no-code tool. It assumes you already use a coding agent and want it to behave like a disciplined engineer instead of an eager one.

Credit

Created by Jesse Vincent (obra)· MIT licensed, free on GitHub

ProfitizeOS did not build this and does not host it. You get a link straight to the original.

Free. Two fields, then straight to the resource and setup guide.

Common questions

Is Superpowers free?

Yes. It is MIT licensed and free on GitHub, so you can read every skill, change any of them, and use it commercially.

How do I install Superpowers in Claude Code?

Run /plugin install superpowers@claude-plugins-official in Claude Code. You can also register the author's own marketplace with /plugin marketplace add obra/superpowers-marketplace and install from there. Restart your agent afterwards.

Does it work with agents other than Claude Code?

Yes. It supports Codex, Cursor, Gemini CLI, GitHub Copilot CLI, OpenCode and several others. Installation differs per harness, so install it separately for each one you use.

What does it actually change about how the agent works?

It changes the order of operations. The agent agrees a specification with you first, then writes an implementation plan for you to approve, and only then writes code, working test-first through subagents rather than jumping straight to an answer.

More on this topic: AI Automation