StackGuide AI
Guides / Builders & agents / Superpowers

How to Install and Use Superpowers

Updated October 31, 2026License: MITโœ“ Checked against official docs

How to Install and Use Superpowers

This guide is based on the open-source project Superpowers by obra, licensed under MIT. StackGuide AI is not affiliated with Superpowers.

Superpowers is a set of composable skills and initial instructions for a coding agent, built around a software development methodology that the README describes as one that works. In practice, it turns a vague request into a design discussion, a written plan of small tasks, implementation by subagents or inline, test-driven development, and a code review before anything is merged. As of October 31, 2026, its GitHub repository has 309,850 stars.

What Superpowers is, and who it is for

The key idea is that the skills are mandatory workflows, not suggestions. The README says the agent checks for relevant skills before any task. That discipline is the whole point: without it, an agent will happily skip the design, skip the tests, and present finished-looking code that nobody reviewed.

It suits developers who already use a coding agent for real features and want the agent's process to be repeatable: brainstorm first, plan in bite-sized tasks, build on an isolated branch, test first, review between tasks, and finish the branch deliberately. It is not a model, a hosted service, or a code generator on its own; it is instructions and skills that steer an agent you already have.

Prerequisites

  • A supported coding agent: the README documents installation for Claude Code, Antigravity, Codex, Cursor, Devin, Factory Droid, Gemini CLI, GitHub Copilot CLI, Grok Build, Kimi Code, OpenCode, Pi, Qwen Code, Hermes Agent and Muse, among others.
  • A Git repository for real work: the workflow creates isolated workspaces on new branches and finishes branches deliberately, so it fits a Git project better than a folder of loose files.
  • Willingness to answer questions first: the brainstorming skill draws out what you actually want and why, and plays its understanding back for correction. If you skip that stage, you lose most of the benefit.

Step 1: Install Superpowers for your agent

Installation differs by harness, and if you use more than one, install Superpowers separately for each one. For Claude Code, use the official marketplace:

/plugin install superpowers@claude-plugins-official

Or register the Superpowers marketplace and install from it:

/plugin marketplace add obra/superpowers-marketplace

/plugin install superpowers@superpowers-marketplace

For other agents, follow the README's section for that tool. Examples:

agy plugin install https://github.com/obra/superpowers

That command is for Antigravity. In the Codex app, use the Plugins sidebar and add Superpowers from the Coding section; in Codex CLI, open /plugins, search for Superpowers, and select Install Plugin; in Cursor Agent chat, use /add-plugin superpowers or search the plugin marketplace. Check the README for your agent rather than guessing a command for a tool not listed here.

Step 2: Follow the basic workflow on a real task

Start with something small but real, and let the skills fire in order:

  1. Brainstorming. Before building anything, describe what you want and why. Small changes get agreed in chat; projects get a written design document a builder could plan from.
  2. Using Git worktrees. After design approval, the agent creates an isolated workspace on a new branch, runs project setup, and verifies a clean test baseline.
  3. Writing plans. With an approved design, the plan breaks work into bite-sized tasks, each with exact file paths, complete code and verification steps. The README describes tasks of two to five minutes each.
  4. Subagent-driven development or executing plans. Either dispatch a fresh subagent per task with a review after each, or implement every task inline in the current session with one fresh review of the whole branch at the end.
  5. Test-driven development. During implementation, the cycle is red, green, refactor: write a failing test, watch it fail, write minimal code, watch it pass, commit. Code written before tests is deleted.
  6. Requesting code review. Between tasks, review against the plan and report issues by severity. Critical issues block progress.
  7. Finishing a development branch. When tasks complete, verify tests, present the options of merge, pull request, keep or discard, and clean up the worktree.

Beyond that sequence, the skills library covers systematic debugging, verification before completion, dispatching parallel agents, receiving code review, and writing your own skills. Explore those only after the basic workflow feels natural.

Common problems and fixes

ProblemFix
The agent skips the workflowAsk it to diagnose the session: the README says to ask your coding agent to figure out what went wrong with Superpowers in this session, which invokes the diagnosing-superpowers skill.
A skill fires when it should not, or stays silentUse the same diagnosing skill, and name the earlier session by its ID if you are examining one after the fact. It reports what happened with line-level evidence.
It works in one agent but not anotherInstall Superpowers separately for each harness. An install in Claude Code does not carry over to Cursor, Codex or any other tool.
Plans are vagueGo back to brainstorming. A plan can only be as precise as the design it came from, and the README's standard is exact file paths, complete code and verification steps per task.
Tests were written after the codeThat breaks the test-driven skill's rule. The skill deletes code written before tests, so restart that task test-first rather than patching around it.

FAQ

Is Superpowers a coding model? No. It is a skills framework and methodology that steers a coding agent you already use. The model behind the agent is unchanged.

Do I have to use every skill? No, but the workflow skills are meant to fire when their situation arises, not when you remember them. The value is in the sequence, especially brainstorming before building and review before merging.

Can I report a broken session usefully? Yes. The diagnosing skill can package a scrubbed bundle for a bug report, so you do not have to paste a raw transcript into an issue.

Who builds Superpowers? The README credits Jesse Vincent and the team at Prime Radiant, with community support on Discord and issues on GitHub.

Licence and attribution

Superpowers is licensed under MIT. In plain terms, and not as legal advice: you may use, copy, modify and distribute it, including commercially, provided you keep the copyright and licence notices with the code. MIT is a permissive licence with no copyleft condition, so it does not require you to open-source a project that uses it, but the notice requirement still applies.

What to try next

Take one small feature through the full sequence, from brainstorming to a finished branch, and read the review the agent produces between tasks. If the review catches something you would have missed, the workflow has already paid for itself.

This guide was checked against the official README and repository as of October 31, 2026. StackGuide AI did not run this tool for this guide, so this guide makes no claim based on running the tool; the commands below are the ones the official README publishes.

Sources

Recommended kit

Private AI Starter Pack โ€” $29 one-time

Ollama + AnythingLLM, pre-wired with Docker, plus a step-by-step guide and install, backup and update scripts.

Buy โ€” $29 Details →

Checkout by Creem.

Newsletter

This Week in Open Source AI

A weekly roundup of open-source AI projects worth your attention.

  • Short takes on the projects, checked against their repos
  • A licence note for each project, because the licence is part of the decision
  • Notice when a new guide is published