Claude Code superpowers

Superpowers

A high-signal Claude Code extension project for skills, workflow upgrades, and agent coding habits.

Last updated: 2026-07-08

What it does

Superpowers extends Claude Code with reusable skills and stronger workflow patterns, so repeated coding tasks can be packaged instead of handled as one-off prompts.

Best fit

Claude Code users who want battle-tested workflow extensions.

Capabilities

  • Starts with brainstorming instead of jumping into code: the agent asks what you are trying to build, extracts a spec, and shows it in reviewable chunks.
  • After design approval, it writes implementation plans with small tasks, exact file paths, code expectations, and verification steps.
  • Ships a skills library: TDD, systematic debugging, verification before completion, brainstorming, writing plans, executing plans, code review, git worktrees, and finishing branches.
  • Supports subagent-driven development: fresh subagents work through tasks and review for spec compliance and code quality.
  • Can be installed across multiple harnesses including Claude Code, Codex App, Codex CLI, Cursor, OpenCode, Antigravity, Kimi Code, Pi, and Copilot CLI.
  • Workflow skills can trigger automatically, so the agent checks for the relevant process before starting a task.

Use cases

  • Add reusable skills to Claude Code projects.
  • Study how a popular Claude Code extension structures workflows.
  • Use brainstorming when the user has an idea but no clear spec yet.
  • Use writing-plans before a multi-file change so every task has files and verification.
  • Use test-driven-development for risky logic so the agent writes failing tests first.
  • Use using-git-worktrees when the agent needs an isolated branch and clean baseline.
  • Use requesting-code-review between tasks to block critical issues before continuing.

Plugin and skill directory