Claude Code Superpowers: What It Is and How to Install It (2026)

"Claude Code superpowers" is real search demand pointing at one specific repo: using-superpowers, with 167,308 installs and 220,998 GitHub stars (skills.sh registry, obra/superpowers) - one of the highest-starred entries in the AI agent tooling category.

Key takeaways

• using-superpowers has 167,308 installs and 220,998 GitHub stars (skills.sh registry, GitHub) from the obra/superpowers repo.
• The bundle covers brainstorming, systematic-debugging, writing-plans, requesting-code-review and test-driven-development as coordinated sub-skills, not one monolithic prompt.
• A community port exists too - workflow-runner (662 installs) reimplements the workflow in a separate repo.
• developing-claude-code-plugins (99 installs) is a companion skill from the same author for building further Claude Code plugins.
• Skillselion currently indexes 65,651 skills tagged to the "build" stage (skills.sh registry) - superpowers sits in the agent-tooling slice of that.

What does "Claude Code superpowers" actually refer to?

It's the common name for the obra/superpowers skill collection - a structured set of sub-skills covering the full build loop: brainstorming an approach, writing a plan, implementing with systematic debugging discipline, running test-driven development, and requesting a structured code review before merging.

What is the best skill for the superpowers workflow in Claude Code?

using-superpowers is the canonical install at 167,308 installs and 220,998 stars (skills.sh registry, GitHub). Install it with:

npx skills add obra/superpowers --skill using-superpowers

For teams extending the workflow into their own plugins, developing-claude-code-plugins (99 installs) from the same author is the companion skill. See the full AI agent skills shortlist.

Is there a lighter-weight alternative?

Yes - for teams that don't need the full multi-skill bundle, single-purpose alternatives exist in the catalog: test-driven-development and systematic-debugging are addressed within the same repo as narrower entry points, and community ports like workflow-runner (662 installs) reimplement a subset of the workflow independently.

How does this differ from a generic agent-tooling skill?

Most agent-tooling skills add a single capability (a tool call, an API wrapper). Superpowers is a workflow methodology - it changes how Claude Code approaches a task (plan first, debug systematically, review before merge) rather than adding one new function call.

A complete starter stack

• using-superpowers - the full brainstorm-to-review workflow bundle
• developing-claude-code-plugins - extend the workflow into custom plugins
• workflow-runner - a lighter community alternative for a subset of the workflow

Common pitfalls

• Installing a same-named community fork expecting the original obra/superpowers feature set - check the repo before installing.
• Treating superpowers as a single skill rather than a coordinated bundle - some sub-skills (like code review) only trigger at specific workflow stages.
• Skipping the planning/brainstorming sub-skill and going straight to implementation, which defeats the workflow's main benefit.

Install the full bundle from obra/superpowers first - narrow to individual sub-skills only once the workflow proves useful for your team.

Common questions