Superpowers
obra/superpowers
A skills pack and development method that makes a coding agent design, plan, test and review before it ships code.
Superpowers is a free, MIT-licensed pack of 15 agent skills by Jesse Vincent and Prime Radiant. It gives a coding agent a fixed way of working: agree a design, write a plan, build with test-driven development, then review. It installs as a plugin in Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot CLI and 10 other agents, and adds a session-start hook so the skills trigger without being called.
Facts checked on against the sources listed at the end of this page.
- GitHub stars
- 294,330
- Skills in the pack
- 15
- Agents supported
- 15named in the README
- Instruction size
- 25.8k wordsacross all SKILL.md files
How to install Superpowers
| Agent | Command | Copy |
|---|---|---|
| Claude Code | From Anthropic's official plugin marketplace. | |
| Codex | In the Codex CLI, run /plugins, search for superpowers and select Install Plugin. In the Codex app, open Plugins in the sidebar and add Superpowers from the Coding section. | |
| Cursor | Run it in Cursor Agent chat. | |
| Gemini CLI | | |
| GitHub Copilot CLI | |
Commands for 10 more agents
| Antigravity | | |
|---|---|---|
| Devin CLI | | |
| Factory Droid | | |
| Grok Build CLI | | |
| Kimi Code | | |
| OpenCode | Tell OpenCode to fetch and follow the instructions in .opencode/INSTALL.md in the repo. | |
| Pi | | |
| Qwen Code | Pick superpowers when prompted. | |
| Hermes Agent | Restart active Hermes sessions after installing. | |
| Muse | |
Commands as given in the README on 2 October 2026.
What is inside Superpowers
| Skill | What it makes the agent do |
|---|---|
| brainstorming | Runs before any code is written. Refines a rough idea through questions, explores alternatives and saves a design document. |
| using-git-worktrees | Creates an isolated workspace on a new branch after the design is approved and checks the tests pass first. |
| writing-plans | Breaks the approved design into tasks of 2 to 5 minutes, each with exact file paths, code and verification steps. |
| subagent-driven-development | Sends each task to a fresh subagent and reviews the result in two stages: against the spec, then for code quality. |
| executing-plans | Runs every task in the current session with one review of the whole branch at the end. |
| test-driven-development | Enforces red, green, refactor: write a failing test, see it fail, write the minimum code, see it pass. |
| requesting-code-review | Reviews work against the plan between tasks and reports issues by severity. Critical issues block progress. |
| systematic-debugging | A four-phase process for finding the root cause of a bug before changing code. |
| verification-before-completion | Checks that a fix works before the agent says it is done. |
| finishing-a-development-branch | Verifies tests, offers merge, pull request, keep or discard, and cleans up the worktree. |
The other five skills are receiving-code-review, dispatching-parallel-agents, diagnosing-superpowers, writing-skills and using-superpowers.
What Superpowers runs on your machine
| Hooks | Yes. A session-start hook runs when a session starts, is cleared or is compacted, and loads the using-superpowers instructions so the agent checks for a relevant skill before each task. |
|---|---|
| Bundled scripts | Yes. The brainstorming skill ships scripts that start and stop a local server for its optional visual companion, and systematic-debugging ships a find-polluter.sh script. |
| Network calls | Yes. By default the optional visual companion in brainstorming loads the Prime Radiant logo from the maker's website, which tells the maker which Superpowers version is in use. Set SUPERPOWERS_DISABLE_TELEMETRY to any true value to turn it off. |
| API keys needed | None named in the README. |
These are facts read from the repo, not a security rating. Read the files before you install any skill.
How much context Superpowers uses
| All 15 skills | 25,796 words across every SKILL.md file |
|---|---|
| Largest skill | subagent-driven-development, 4,871 words |
| Loaded at start | Only each skill's name and description. The full text loads when a task needs that skill. |
Word counts are measured from the repo every day. Reference files a skill loads on demand are not counted.
How Superpowers works
Superpowers changes what a coding agent does when you ask it to build something. Before writing code, the agent asks what you are trying to do, turns the answers into a design and shows it to you in short sections. Once you approve, it writes an implementation plan made of small tasks and works through them, either with a fresh subagent per task or inline in one session.
The skills trigger automatically. A session-start hook loads a short set of instructions that tells the agent to check for a relevant skill before any task, so the workflow is mandatory, not a suggestion. The README says the same repo and skills work across every supported agent, but each agent needs its own install.
A good fit for
- Teams that want the agent to agree a design and a plan before it writes code
- Long tasks where an agent drifts or skips tests without a fixed process
- People who use more than one coding agent and want the same workflow in each
- Projects that already work test-first
Skip it if
- Quick one-file edits, where the design and planning steps add time
- People who want to pick single skills: the pack is built to be used as a whole method
- Setups that do not allow session hooks or outbound requests, unless you turn the telemetry off
Is Superpowers still maintained?
| Last commit | , 1 week ago |
|---|---|
| Latest release | v6.4.2, 25 September 2026 |
| Stars gained in 7 days | Shown after 7 days of tracking |
| Stars gained in 30 days | Shown after 30 days of tracking |
| Open issues and pull requests | 295 |
| Forks | 26,313 |
Tools Superpowers works with
Questions about Superpowers
Does Superpowers work with Codex and Cursor?
Yes. The README gives install steps for Claude Code, Codex (app and CLI), Cursor, Gemini CLI, GitHub Copilot CLI and 10 other agents. You install it separately in each agent you use.
Is Superpowers free?
Yes. It is open source under the MIT license. Prime Radiant offers paid commercial support for companies.
Is Superpowers safe to install?
The code is open, so every skill, hook and script can be read first. It adds a session-start hook, and the optional visual companion loads a logo from the maker's website unless you set SUPERPOWERS_DISABLE_TELEMETRY.
How do I update Superpowers?
It depends on the agent, and updates are often automatic. The README gives update commands for Devin CLI, Gemini CLI, Qwen Code and Muse.
What is the difference between subagent-driven-development and executing-plans?
Both carry out a written plan. subagent-driven-development sends each task to a fresh subagent with a review after each one, which the README calls the most thorough. executing-plans does every task in the current session with one review at the end, which it calls the cheapest.