superpowers
An agentic skills framework & software development methodology that works.
Superpowers is a plugin for AI coding agents like Claude Code, Cursor, and Codex that enforces a disciplined workflow, brainstorm a spec first, break it into tasks, implement with test-driven development, then run a two-stage code review.
Superpowers is, in the README's own words, "a complete software development methodology for your coding agents, built on top of a set of composable skills and some initial instructions that make sure your agent uses them." A coding agent here is an AI assistant that writes code on your behalf inside a tool like Claude Code, Codex CLI, Cursor, Gemini CLI, or GitHub Copilot CLI. Superpowers is a plugin that ships into that agent and changes how it behaves on a coding task. The way it works, per the README, is that when you fire up your agent on a build task it does not jump straight into writing code. Instead it starts a brainstorming phase, asks what you are really trying to do, and teases a spec out of the conversation. After you sign off, the agent produces an implementation plan broken into small tasks (each with file paths, code, and verification steps), then launches a "subagent-driven-development" process where fresh subagents work through each task and a two-stage review checks spec compliance and code quality. The methodology emphasizes test-driven development (RED-GREEN-REFACTOR), YAGNI, and DRY, with companion skills for git worktrees, code review, finishing a branch, systematic debugging, and writing new skills. Someone would use Superpowers when they want their coding agent to follow a disciplined workflow — design first, plan, test-first implementation, review — instead of cowboy-coding. Installation differs per harness: Claude Code via the plugin marketplace, Codex CLI and App via /plugins, plus separate flows for Factory Droid, Gemini CLI, OpenCode, Cursor, and Copilot CLI.
Where it fits
- Add structured spec-writing and planning to your Claude Code or Cursor agent before it writes any code.
- Enforce test-driven development on an AI coding assistant using a built-in RED-GREEN-REFACTOR workflow.
- Use the code review skill to have a fresh AI subagent check spec compliance and code quality after each task.
- Set up git worktrees with Superpowers to run multiple AI coding tasks in parallel without conflicts.