Superpowers skills for Claude Code
Superpowers is a community set of skills by Jesse Vincent and Prime Radiant that gives a coding agent a fixed method: brainstorm the design, write a plan, build test-first, review, then finish the branch.
/plugin install superpowers@claude-plugins-officialSuperpowers describes itself as a complete software development method for coding agents, built from skills that trigger on their own. The author is Jesse Vincent (GitHub user obra), working with Prime Radiant, and the license is MIT. It is listed in Anthropic's official Claude Code plugin marketplace, and maintained by its authors. For the basics of skills, read how Claude Skills work.
What the skills do
The README lays out the default workflow, one skill per step:
- brainstorming: before any code, the agent asks what you are trying to do, explores alternatives and shows the design in short sections for approval.
- using-git-worktrees: a separate branch and workspace, with a clean test baseline.
- writing-plans: tasks of two to five minutes each, with exact file paths and verification steps.
- subagent-driven-development or executing-plans: a fresh subagent per task with a review after each, or everything inline with one review at the end.
- test-driven-development: write a failing test, watch it fail, write the minimum code, watch it pass. Code written before its test gets deleted.
- requesting-code-review: a review against the plan, with critical issues blocking progress.
- finishing-a-development-branch: tests verified, then merge, pull request, keep or discard.
Other skills cover systematic debugging in four phases, verification before declaring a fix, parallel agents and writing new skills.
When to use it
- Features big enough to plan, where an agent tends to drift.
- Teams that want test-first work by default.
- Longer unattended runs. The README says the agent often works for a couple of hours on its own once you approve the plan.
For one quick fix it is a lot of process. If you only want the questioning step, grill-me is lighter, and the bundled /code-review skill covers review alone. To check UI work in a browser, add webapp-testing.
How to install it
Claude Code, from Anthropic's official marketplace:
/plugin install superpowers@claude-plugins-official
Or from the author's own marketplace:
/plugin marketplace add obra/superpowers-marketplace
/plugin install superpowers@superpowers-marketplace
Codex: in the Codex app or CLI, open Plugins and install Superpowers from the Coding section. The README also gives steps for Cursor, Gemini CLI, GitHub Copilot CLI and other agents. Our Claude Code vs Codex comparison covers the two main options.
Example request
I want to add [feature] to [project]. Users need to [what they do]. Constraints: [stack, deadline, what must not change]. Let's brainstorm the design before any code.
Limits
- The workflows are mandatory by design, per the README. Expect questions and a plan before code, even when you are in a hurry.
- Subagent-driven work uses more tokens than inline execution. The README calls inline execution the cheaper option.
- The brainstorming skill's optional visual companion loads a logo from Prime Radiant's site, which reports the Superpowers version. Set
SUPERPOWERS_DISABLE_TELEMETRYto turn it off; it also honors Claude Code's telemetry opt-outs. - The project generally does not accept new skills from contributors.
Sources
- obra/superpowers README — GitHub, accessed October 2026
- Superpowers skills folder — GitHub, accessed October 2026