Superpowers Setup: Give Your AI Coding Agent a Real Workflow

Superpowers is a free, open source skills framework that bolts a complete, disciplined software development methodology onto your coding agent, and it is one of the loudest repositories on GitHub trending today, adding roughly 740 stars in a day on top of more than 250,000 total. Instead of letting an AI agent sprint straight into code, Superpowers makes it stop, tease a real spec out of you, write a bite sized plan, and build against tests. Setup is genuinely fast: one plugin install command for your agent, and the skills start firing on your very next message. This is our daily Tutorials pick, a real trending tool set up for real.

• Superpowers is MIT licensed and installs as a plugin, with official support for Claude Code, Codex, Cursor, Antigravity, Factory Droid, GitHub Copilot CLI, Kimi Code, OpenCode, and Pi.
• It ships a library of composable skills that trigger automatically: brainstorming, planning, git worktrees, test driven development, code review, and subagent driven development.
• The core idea is process over vibes: your agent refuses to guess, writes a plan an enthusiastic junior could follow, and enforces red green refactor TDD.
• Install is per harness: if you use more than one coding agent, you install Superpowers separately for each, and the current release is v6.1.1.

What is Superpowers and why is it trending?

Superpowers is built by Jesse Vincent and the team at Prime Radiant, and it describes itself as a complete software development methodology for coding agents, layered on top of a set of composable skills plus a small bootstrap that makes sure the agent actually uses them. Under the hood it is mostly shell and markdown: skill files that describe when and how the agent should brainstorm, plan, test, and review, wrapped in a plugin that each supported harness loads at session start. The repository has crossed 250,000 stars and 22,000 forks since its October 2025 debut, and the daily spike is easy to explain. Coding agents got fast enough to write real code this year, but most of them still behave like an overeager intern: they guess at requirements, skip tests, and declare victory early. Superpowers is the most talked about attempt to impose a senior engineer's process on that intern, and it works across nearly every popular agent, so the audience is enormous.

RelatedDesktop Commander Setup: Give Claude Terminal Control

How do you install Superpowers on Claude Code?

Claude Code is the reference harness, and Superpowers is in Anthropic's official plugin marketplace, so installation is a single slash command typed into your agent:

# Claude Code: install from the official marketplace
/plugin install superpowers@claude-plugins-official

If you would rather pull from the project's own marketplace, which also carries related plugins, register it first and then install:

# Claude Code: install from the Superpowers marketplace
/plugin marketplace add obra/superpowers-marketplace
/plugin install superpowers@superpowers-marketplace

That is the entire setup. The skills trigger automatically, so you do not add flags or memorize commands. The next time you ask your agent to build something, it steps back and starts asking what you are really trying to do instead of writing code.

How do you install it on Cursor, Codex, and other agents?

Superpowers installs per harness, so if you switch between agents you repeat the step for each one. The commands come straight from the project README:

# Cursor: in Agent chat
/add-plugin superpowers

# Codex CLI: open the plugin search, then find Superpowers
/plugins

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

# Factory Droid
droid plugin marketplace add https://github.com/obra/superpowers
droid plugin install superpowers@superpowers

# GitHub Copilot CLI
copilot plugin marketplace add obra/superpowers-marketplace
copilot plugin install superpowers@superpowers-marketplace

# Pi
pi install git:github.com/obra/superpowers

Kimi Code and OpenCode are supported too, each through their own plugin manager, and the README links per agent docs for the fiddlier ones. The through line is the same everywhere: install once per agent, then forget it exists.

What does the Superpowers workflow actually do?

The value is in the sequence of skills that fire as you build. It opens with brainstorming, a Socratic pass that refuses to write code until it has teased a spec out of the conversation and shown it to you in chunks short enough to read. Once you approve the design, using-git-worktrees spins up an isolated branch and verifies a clean test baseline, then writing-plans breaks the work into two to five minute tasks, each with exact file paths and verification steps. Implementation runs through subagent-driven-development, which dispatches a fresh subagent per task with a two stage review for spec compliance and then code quality, while test-driven-development enforces red green refactor and deletes any code written before its test. Between tasks, requesting-code-review reports issues by severity and blocks progress on critical ones, and finishing-a-development-branch handles the merge or pull request decision at the end. Debugging skills like systematic-debugging and verification-before-completion round it out. Because everything triggers automatically, the agent can run autonomously for a couple of hours without wandering off the plan you approved.

RelatedClaude Video Setup: Let Claude Watch Any Video

How does it compare with the alternatives?

The honest comparison is not Superpowers versus a rival product; it is Superpowers versus the pile of custom instructions and rule files most serious agent users cobble together themselves. It packages that discipline into a maintained, tested plugin that works the same way across nine different agents, which is exactly the chore people are tired of redoing.

What are the gotchas before you rely on it?

Four things worth knowing. First, Superpowers is a methodology, not a model upgrade: it makes a capable agent disciplined, but it cannot make a weak one smart, and the output still depends on the underlying coding agent. Second, install is genuinely per harness, so a setup that works in Claude Code does nothing for your Cursor sessions until you install it there too. Third, there is optional telemetry: by default the brainstorming skill's visual companion loads a logo from Prime Radiant's servers, carrying only the Superpowers version and no project data, and you turn it off by setting SUPERPOWERS_DISABLE_TELEMETRY to any true value or by using Claude Code's existing telemetry opt outs. Fourth, this is an opinionated process: it wants spec sign off, git worktrees, and strict TDD, so if your workflow fights any of those, expect friction. None of these are dealbreakers, but they are the difference between a smooth adoption and a surprised one.

Our take

Superpowers is less a tool than an argument, and the argument is correct: the bottleneck on agent coding stopped being raw capability and became discipline. The reason a plugin of shell scripts and markdown can pull 250,000 stars is that it names the thing every heavy agent user already discovered, that the difference between a useful agent and a dangerous one is whether it plans, tests, and reviews before it declares victory. Packaging that as a one command install across nine harnesses is genuinely useful, and the price ends the debate. The thing to watch is durability. Riding on top of agents that change every week is a hard place to maintain a methodology, and the commercial split will test whether the free edition stays first class. Today, if you run any supported coding agent and you are tired of babysitting it, this is the fastest way to give it a real process.

Original analysis by GenZTech. Tool documentation: obra/superpowers on GitHub.