Superpowers
Superpowers
Helps Claude Code work more methodically; intended for people already comfortable with programming.
A community skill set for Claude Code: planning, test-first development, systematic debugging, code review.
Overview
Superpowers is a software development methodology for coding agents, built from composable skills. The project is developed by Jesse Vincent and the Prime Radiant team and released under the MIT license.
What sets it apart is that Superpowers makes the agent follow a disciplined process: ask questions to clarify requirements and design first, write a detailed plan, and only then implement using TDD with subagents and code review. The skills activate automatically, so you do not need to remember special commands while working.
The skill set supports many tools: Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot CLI, Antigravity, Hermes Agent and many other agents. For dev teams, it is a quick way to have AI work like an engineer with a process, instead of writing code before the requirements are clear.
Highlights
Design first, code later
The brainstorming skill asks follow-up questions to clarify the idea and saves a design document before any code is written.
Plans in small pieces
writing-plans breaks work into 2 to 5 minute tasks, with file paths, code and concrete verification steps.
Mandatory TDD
test-driven-development applies the red-green-refactor cycle, including deleting code written before its test.
Subagents and review
Each task can be given to a fresh subagent with review; serious issues block progress.
Multi-agent support
Install guides for Claude Code, Codex, Cursor, Gemini CLI, Copilot CLI, Antigravity, Hermes Agent and many other platforms.
What it's for
Building new features with control
Dev teams use brainstorming and writing-plans to settle requirements with the client before letting AI write code.
Systematic bug fixing
Use the systematic-debugging skill to find the root cause instead of patching piecemeal.
Safe parallel work
using-git-worktrees creates a separate working branch so the AI does not touch running code.
Standardizing code review
requesting-code-review and receiving-code-review make sure every change is checked against the plan.
How to use it
- 1
Open Claude Code
Start Claude Code in the project where you want to apply Superpowers.
- 2
Install from the official marketplace
Install the Superpowers plugin from Claude Code's official marketplace.
$/plugin install superpowers@claude-plugins-official - 3
Or use the dedicated marketplace
Add Superpowers' own marketplace, then install the plugin.
$/plugin marketplace add obra/superpowers-marketplace - 4
Install the plugin from the dedicated marketplace
After adding the marketplace in the previous step, run this command to install Superpowers.
$/plugin install superpowers@superpowers-marketplace - 5
Restart the session
After installing, restart any open sessions so the hooks take effect.
- 6
Start by describing your idea
State what you want to build; the agent will ask follow-up questions and draft a design before writing code.
Video guides
Claude Code + SUPERPOWERS = The End of Vibe Coding? (Full Tutorial)
Eric Tech
The Claude Code Plugin Every Developer Must Learn (Superpowers)
GritAI Studio
Superpowers + Claude Code: Full Test & Honest Review
AI Unleashed
Real-world example
Suppose a small software startup in Da Nang needs to add appointment booking to a spa app. The developer installs Superpowers in Claude Code and describes the requirements. The agent asks about time slots, cancellations and notifications, saves a design, then breaks the plan into many small tasks. After the developer approves, the agent creates a separate worktree, writes tests first for each part, assigns work to subagents and reviews it before proposing a merge or creating a PR.
Tips & notes
Review the design carefully
Output quality depends on the design and plan, so take time to answer the agent's questions.
Choose how to execute
subagent-driven-development is the most thorough; executing-plans runs smoothly and is cheaper, with one final review.
Turn off telemetry if needed
Set the SUPERPOWERS_DISABLE_TELEMETRY variable to stop the logo download that includes the Superpowers version.
Self-diagnose
When the process runs wrong, ask the agent to investigate the issue with superpowers to trigger the diagnosing-superpowers skill.
FAQ
Is Superpowers free?
The project is open source under the MIT license; AI usage costs depend on the agent you run.
Do I have to type commands to invoke each skill?
No, the skills activate automatically; the agent checks for relevant skills before each task.
How do I install it if I use multiple agents?
The README has separate install instructions for each platform; if you use several tools, install for each one.
Does it accept new skill contributions?
The README says the project generally does not accept new skills, and any change must work on all supported agents.
Useful links
Want to apply this in your business?
2dm looks at your workflows, picks the right tools and sets them up with your team. The first call is free.
Same topic
Anthropic Skills
4.5
Anthropic's official repository of example skills for Claude: each skill is a folder of instructions and resources that helps Claude do one specific kind of task well.
Agent Skills documentation
4.0
Explains what a skill is, the SKILL.md folder structure, and how Claude loads skills automatically when needed.
Skills in Claude Code
4.0
How to create, install and use your own skills for Claude Code, on your machine or per project.