Is Superpowers made by Anthropic?

No. It is built by Jesse Vincent (obra on GitHub) and Prime Radiant. It is distributed through Anthropic's official Claude plugin marketplace, which is a listing, not authorship — support comes from the project's Discord and issue tracker.

Yes. The code is MIT-licensed and the plugin costs nothing. Prime Radiant separately sells commercial support and enterprise tooling, which funds the project.

Does it only work with Claude Code?

No. The same repo documents installs for roughly a dozen harnesses, including Codex, Cursor, Gemini CLI, GitHub Copilot CLI, and OpenCode. Each harness needs its own separate install.

Does it send any data anywhere?

One thing: the brainstorming skill's visual companion loads a logo from Prime Radiant's site, which reports your Superpowers version — nothing about your project or prompts. Setting SUPERPOWERS_DISABLE_TELEMETRY to any true value turns it off.

Skill review · by Jesse Vincent (obra) and Prime Radiant

Claude Superpowers Skill

The Claude superpowers skill this search points to is Superpowers, an open-source development methodology from Jesse Vincent and the team at Prime Radiant, shipped as a plugin carrying 14 composable skills. It sits at 280,319 GitHub stars as of September 2026, with v6.3.0 released in August 2026 and commits landing the day before this review, so it is both widely used and actively maintained. It installs into Claude Code from Anthropic's official plugin marketplace with one command, and the same repo installs into Codex, Cursor, Gemini CLI, and roughly a dozen other coding agents. The one thing to know before installing: the skills trigger automatically and the workflow is mandatory by design — brainstorming before code, written plans, red-green TDD, subagent execution with two-stage review. That discipline is the product. If you want Claude to just write the function, this framework will slow you down on purpose. It is not an Anthropic product, and by default it reports its version to Prime Radiant via a logo load, which one environment variable disables.

View the sourceMIT licensed · 280,319★ · facts checked September 1, 2026

By Robert MacKelfresh

Founder, Automation Squad · September 1, 2026

Install

1In Claude Code, run /plugin install superpowers@claude-plugins-official — it is listed on Anthropic's official plugin marketplace.
2Alternative source: /plugin marketplace add obra/superpowers-marketplace, then /plugin install superpowers@superpowers-marketplace.
3Start a new session. A session-start hook injects the using-superpowers bootstrap; skills then trigger automatically — there is nothing to invoke by hand.
4Using other agents too? Install separately per harness: Codex via /plugins, Cursor via /add-plugin superpowers, Gemini CLI via gemini extensions install https://github.com/obra/superpowers. Installs do not carry across harnesses.
5Optional: set SUPERPOWERS_DISABLE_TELEMETRY=1 to stop the version-reporting logo load (it also honors DISABLE_TELEMETRY and CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC).

What it does

Design before code: The brainstorming skill activates the moment you describe something to build. It refuses to jump into code, refines the idea through Socratic questioning, presents the design in short reviewable sections, and saves a design document you sign off on.
Implementation planning: writing-plans breaks approved designs into tasks of 2-5 minutes each, with exact file paths, complete code, and verification steps — written, per the README, for 'an enthusiastic junior engineer with poor taste' to follow safely.
Subagent execution: subagent-driven-development dispatches a fresh subagent per task with two-stage review (spec compliance, then code quality); executing-plans is the batch alternative with human checkpoints. Autonomous runs of a couple hours without deviating from the plan are normal.
Testing and debugging discipline: test-driven-development enforces the RED-GREEN-REFACTOR cycle and deletes code written before its test existed. systematic-debugging adds a 4-phase root-cause process, and verification-before-completion blocks 'done' claims until proven.
Branch lifecycle: using-git-worktrees creates an isolated workspace per feature and verifies a clean test baseline; finishing-a-development-branch runs tests, offers merge/PR/keep/discard, and cleans up the worktree.

Common questions

Is Superpowers made by Anthropic?: No. It is built by Jesse Vincent (obra on GitHub) and Prime Radiant. It is distributed through Anthropic's official Claude plugin marketplace, which is a listing, not authorship — support comes from the project's Discord and issue tracker.
Is Superpowers free?: Yes. The code is MIT-licensed and the plugin costs nothing. Prime Radiant separately sells commercial support and enterprise tooling, which funds the project.
Does it only work with Claude Code?: No. The same repo documents installs for roughly a dozen harnesses, including Codex, Cursor, Gemini CLI, GitHub Copilot CLI, and OpenCode. Each harness needs its own separate install.
Does it send any data anywhere?: One thing: the brainstorming skill's visual companion loads a logo from Prime Radiant's site, which reports your Superpowers version — nothing about your project or prompts. Setting SUPERPOWERS_DISABLE_TELEMETRY to any true value turns it off.

Before you install

Not an Anthropic product. It is third-party software from Jesse Vincent and Prime Radiant; distribution happens through the official marketplace, but support is a community Discord and a GitHub tracker with 348 open issues.
Phones home by default. The brainstorming skill's optional visual companion loads the Prime Radiant logo from their site, carrying your Superpowers version number. The README is upfront about it and documents the opt-out, but it is on until you turn it off.
The workflow is mandatory, not advisory — the README's own words are 'Mandatory workflows, not suggestions.' For a one-line fix, the brainstorm-plan-TDD pipeline is overhead, and multi-hour autonomous subagent runs consume tokens accordingly.
It is not a single uploadable skill. This is a 14-skill plugin for coding agents; it is not the kind of skill folder you zip and upload to claude.ai's Settings, and it has no use outside a coding harness.
The maintainers generally do not accept contributions of new skills, and any skill change must work across every supported agent — so extending it means forking, not PRing.
A commercial company sits behind it. The core is free MIT, but Prime Radiant sells enterprise support and tooling, and the README carries the sales pitch.
Harness support is uneven at the edges — Hermes, for one, loses the bootstrap after context compaction and needs a fresh session when skills stop triggering.

Facts come from the skill's own repository and docs on the date above, via the GitHub API — not from memory, and never from a sponsor.

Worth comparing

anthropics/skills — Anthropic's first-party skills repo — narrower, mostly document and task skills, no enforced methodology.
Claude Code built-ins — Plan mode, /code-review, and native skills cover pieces of the same loop without adopting a framework.

← All MCP servers & skills