Claude Superpowers: What It Is and How to Install It

Superpowers is an open source Claude Code plugin by Jesse Vincent (obra on GitHub) that installs a complete software development methodology as a set of 14 composable skills, covering brainstorming, planning, test-driven development, and systematic debugging. You install it inside Claude Code with a single command: /plugin install superpowers@claude-plugins-official.

What Is the Superpowers Plugin?

Superpowers is an agentic skills framework for coding agents, described in its own README as "a complete software development methodology for your coding agents, built on top of a set of composable skills." Jesse Vincent, the creator of Request Tracker and co-founder of Keyboardio, announced it on his blog on October 9, 2025, the same day Anthropic launched the Claude Code plugin system.

The core idea: instead of letting Claude jump straight into writing code, Superpowers forces it to step back. The plugin ships skills plus bootstrap instructions that tell Claude to search for a relevant skill before acting and to follow it when one exists. When you ask for a feature, Claude first interviews you to tease out a real spec, then plans, then builds.

The project moved fast after launch. Vincent shipped Superpowers 4 on December 18, 2025, and by June 1, 2026 the plugin had recorded 752,120 installs in the official Claude plugin directory, second only to Anthropic's own frontend-design plugin, according to Composio's install-count roundup. As of July 2026 the GitHub repo sits at 243,000+ stars.

What Superpowers Includes: The 14 Skills

The Superpowers repo ships 14 skills in its skills/ directory as of July 2026, grouped into testing, debugging, collaboration, and meta categories. Each skill is a Markdown instruction set that Claude loads on demand, which keeps the always-on context cost low.

• Testing: test-driven-development, which enforces a strict RED-GREEN-REFACTOR loop.
• Debugging: systematic-debugging and verification-before-completion, so Claude proves a fix works instead of claiming it does.
• Collaboration: brainstorming, writing-plans, executing-plans, dispatching-parallel-agents, subagent-driven-development, requesting-code-review, receiving-code-review, using-git-worktrees, and finishing-a-development-branch.
• Meta: writing-skills and using-superpowers, which teach Claude how to author new skills and how to find the ones it already has.

Because the skills are composable, they chain automatically: brainstorming produces a design document, writing-plans turns it into small tasks, and executing-plans runs those tasks with subagents and review gates. If the SKILL.md format is new to you, our Claude plugins guide explains how plugins, skills, and MCP servers fit together.

How to Install Superpowers

Installing Superpowers in Claude Code takes one command because it is listed in the official plugin marketplace, which Claude Code registers automatically. Run this inside a Claude Code session:

# From the official Anthropic marketplace (already registered)
/plugin install superpowers@claude-plugins-official

# Or from the author's own marketplace, which gets updates first
/plugin marketplace add obra/superpowers-marketplace
/plugin install superpowers@superpowers-marketplace

After installing, run /reload-plugins or restart Claude Code, then start any non-trivial coding request. You do not need to invoke anything manually; the bootstrap instructions make Claude reach for the skills on its own. The README also documents install paths for other agents, including /add-plugin superpowers in Cursor, copilot plugin marketplace add obra/superpowers-marketplace for GitHub Copilot CLI, and a fetch-and-follow install file for OpenCode.

Superpowers is MIT licensed and free. The only costs are indirect: the workflow generates more model turns (planning, subagents, reviews), so token usage per feature is higher than freestyle prompting.

How the Superpowers Workflow Runs

A Superpowers session moves through six phases, per the project README: brainstorming, git worktree setup, planning, development, test-driven development, and code review with branch finalization. Each phase is enforced by its own skill rather than by a monolithic prompt.

1. Brainstorming: Claude interviews you about the problem before writing any code, then presents the emerging spec in short chunks for approval.
2. Git worktrees: work happens on an isolated branch with a clean baseline, so experiments never dirty your main checkout.
3. Planning: the feature is broken into tasks sized at roughly 2 to 5 minutes each, with exact specifications.
4. Development: tasks are executed, often by dispatched subagents, with a two-stage review between steps.
5. TDD: every change follows the RED-GREEN-REFACTOR cycle; the test must fail before the implementation is written.
6. Review and finish: the result is reviewed against the plan, and the branch is merged or handed back cleanly.

The philosophy running through all six phases is "evidence over claims": Claude has to demonstrate that something works instead of asserting it. That pairs well with the habits in our Claude Code best practices guide.

Superpowers vs Individual Skills

Superpowers is an opinionated bundle, while picking individual skills lets you assemble only the process you want; the right choice depends on how much methodology you want imposed on every session. Here is how the three realistic options compare:

The approaches are not mutually exclusive. A common pattern is running Superpowers as the backbone and layering project-specific skills on top; Superpowers even includes writing-skills to help Claude author those. If you would rather start from prebuilt project configurations, Claude Code templates cover that angle, and the plugin directory lists Superpowers alongside 3,600+ other Claude Code plugins.

Limitations and Tradeoffs

Superpowers deliberately slows Claude down, and that is not always what you want. For a one-line bugfix or a rename, the brainstorm-plan-execute pipeline is overhead with no payoff; skip it or answer the brainstorm quickly to move on. Token usage is also structurally higher because planning documents, subagent dispatches, and two-stage reviews all consume model turns.

Two more caveats. First, like any Claude Code plugin, Superpowers runs with your user privileges, so apply the same trust review you would to any third-party plugin. Second, the project iterates quickly, with major revisions like Superpowers 4 landing within months of launch, so pin your expectations to the current README rather than older tutorials. If a structured methodology is more than you need, browse the skills directory and adopt one skill at a time instead.

Related Guides

• Claude Plugins: What They Are and How to Install Them
• Claude Code Templates: Prebuilt Setups Worth Using
• Claude Code Best Practices in 2026
• Claude Code MCP: How to Add and Manage MCP Servers