Skip to content
Superpowers

Superpowers

by Jesse Vincent
MIT
Repository Docs
Featured markdown Development
workflowtddplanningcode-reviewsubagentsgit-worktrees

Summary

An agentic software-development methodology: composable skills that push a coding agent through spec, plan, TDD and review instead of straight into code.

Features

  • Brainstorming skill that extracts a reviewable spec before any code is written
  • Plans broken into 2-5 minute tasks with exact file paths and verification steps
  • Subagent-driven development with two-stage review per task
  • Enforced RED-GREEN-REFACTOR TDD that deletes code written before its test
  • Every run isolated in its own git worktree against a clean test baseline
  • Automatic skill triggering - no commands to remember
  • Installable on 14+ harnesses including Claude Code, Cursor, Codex and Gemini CLI

Install This Skill

Add this skill to your favorite AI agent in a few steps.

Any AI agent

This skill is plain instructions — it works with any assistant that accepts custom instructions or system prompts.

  1. Copy the skill content with the button below.
  2. Paste it into your agent's instruction file or system prompt (for example AGENTS.md, .cursorrules, or a custom instructions field).
  3. Ask the agent to apply the skill whenever the task matches.

Skill Content

Markdown Content

Copy this content and use it with your preferred AI agent

# Superpowers

Superpowers is a complete software development methodology for coding agents, built on composable
skills plus the initial instructions that make sure the agent actually uses them.

The agent checks for a relevant skill before any task. These are mandatory workflows, not suggestions.

## The workflow

1. **brainstorming** - Activates before writing code. Refines rough ideas through questions, explores
   alternatives, presents the design in sections short enough to actually read. Saves a design document.
2. **using-git-worktrees** - Activates after design approval. Creates an isolated workspace on a new
   branch, runs project setup, verifies a clean test baseline.
3. **writing-plans** - Activates with an approved design. Breaks work into bite-sized tasks of 2-5
   minutes each. Every task has exact file paths, complete code and verification steps.
4. **subagent-driven-development** / **executing-plans** - Dispatches a fresh subagent per task with
   two-stage review (spec compliance, then code quality), or executes in batches with human checkpoints.
5. **test-driven-development** - Enforces RED-GREEN-REFACTOR: write a failing test, watch it fail,
   write minimal code, watch it pass, commit. Deletes code written before its test.
6. **requesting-code-review** - Activates between tasks. Reviews against the plan and reports issues by
   severity; critical issues block progress.
7. **finishing-a-development-branch** - Verifies tests, presents merge/PR/keep/discard options, cleans
   up the worktree.

## Philosophy

Plans should be clear enough for an enthusiastic junior engineer with poor taste, no judgement, no
project context and an aversion to testing to follow. The methodology emphasises true red/green TDD,
YAGNI and DRY.

Usage Instructions

Learn how to use this skill with different AI agents.

Claude Desktop

Install from the official Claude plugin marketplace:

/plugin install superpowers@claude-plugins-official

Or register the project's own marketplace first:

/plugin marketplace add obra/superpowers-marketplace
/plugin install superpowers@superpowers-marketplace

Example Usage

Just describe what you want to build. Superpowers activates on its own:

Let's build a rate limiter for the API.

The agent will start by asking questions rather than writing code, present a design for approval, write a task-level plan, then execute it in a git worktree under TDD with review between tasks.

Description

Superpowers is the most-installed agent skills framework on GitHub, and it is less a single skill than an opinionated development process that a coding agent is made to follow.

The premise is that the failure mode of coding agents is not writing bad code, it is writing the wrong code confidently. So the moment the agent notices you are building something, it does not start typing. It steps back and interrogates the idea, teases a specification out of the conversation, and shows it back to you in chunks short enough to actually read and sign off on.

From an approved design it writes an implementation plan broken into tasks of two to five minutes each, with exact file paths, complete code and verification steps — deliberately written to be followable by "an enthusiastic junior engineer with poor taste, no judgement, no project context, and an aversion to testing." Then it runs subagent-driven development: a fresh subagent per task, each reviewed twice, first for spec compliance and then for code quality, with critical issues blocking progress. Test-driven development is enforced rather than suggested — RED-GREEN-REFACTOR, and code written before its test gets deleted.

Two structural choices make the whole thing work. Every piece of work happens in its own git worktree with a verified-clean test baseline, so a long autonomous run cannot quietly corrupt your main checkout. And the skills trigger automatically from context, so there is no command to remember; the agent simply behaves differently. Runs of a couple of hours without deviating from the agreed plan are reported as routine.

It is genuinely harness-neutral. Alongside Claude Code, the project documents installation for Cursor, Codex CLI and the Codex app, Gemini CLI, GitHub Copilot CLI, Devin CLI, Antigravity, Factory Droid, Grok Build CLI, Kimi Code, OpenCode, Pi and Hermes Agent — install it separately per harness if you use more than one.

MIT licensed and maintained by Jesse Vincent and the team at Prime Radiant, who also offer commercial support for enterprise use. Companion repositories carry experimental skills (superpowers-lab) and community-editable ones (superpowers-skills).

Best suited to people who want an agent to work autonomously for long stretches and are willing to spend ten minutes on a spec up front to get it. Overkill for one-line fixes.

Related Skills

Skill: SkillHone

by Tencent

New

Tencent's skill-evolution harness: it rewrites a whole skill folder — SKILL.md, scripts and references together — and lands every decision as a real Git issue, PR and wiki entry you can review.

Development

Skill: OpenSearch Launchpad

by OpenSearch Project

New

Take an OpenSearch search application from requirements to a running cluster — BM25, dense and sparse vectors, hybrid retrieval, agentic search and RAG, with relevance evaluation built in.

Development
2 views

Skill: Auth0 Agent Skill

by Auth0

New

Auth0's official agent skill: a router that detects your framework and intent, then loads the right Auth0 guidance for login, MFA, Organizations, tenant audits, debugging or provider migration.

Development

Skill: Redis Search

by Redis, Inc.

New

Redis' own guidance for FT.CREATE schema design, FT.SEARCH / FT.AGGREGATE / FT.HYBRID, HNSW vector similarity and RAG retrieval pipelines.

Development
1 views
Browse all skills →