Superpowers by Obra

Superpowers turns your AI coding agent into a disciplined software-development partner. Instead of jumping straight into code, it guides the agent through brainstorming, specification, planning, test-driven implementation, code review, and branch completion—automatically.

Built on a library of composable skills, Superpowers works with popular coding agents and IDEs, including Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot CLI, OpenCode, Devin, Kimi Code, Qwen Code, and more. Install it once for your preferred harness and the relevant workflow activates when your agent starts building.

Features

Structured development workflow — Moves from idea to approved design, implementation plan, coding, review, and merge or pull request

Automatic skill triggering — Skills activate based on the task; no manual prompting required

Test-driven development — Enforces a RED-GREEN-REFACTOR cycle: write a failing test, implement minimally, verify, and commit

Detailed implementation plans — Breaks approved designs into small, verifiable tasks with file paths, code, and validation steps

Subagent-driven development — Can dispatch fresh agents per task, inspect their output, and review work as it progresses

Systematic debugging — Provides a four-phase root-cause process and verification before declaring a fix complete

Git worktree support — Creates isolated branches and workspaces for safer parallel development

Code review workflows — Reviews changes against the plan, reports issues by severity, and blocks progress on critical problems

Broad agent compatibility — Offers installation paths for many CLI agents, IDE agents, and plugin marketplaces

Open source — Available under the MIT License

How it works

Brainstorm — The agent asks clarifying questions, explores alternatives, and presents the design in readable sections

Plan — It creates a concrete implementation plan with small, testable tasks

Implement — It follows TDD and either executes tasks inline or coordinates subagents

Review — It checks work against the plan and requests code review between tasks

Finish — It verifies tests, helps choose merge, PR, keep, or discard, and cleans up the worktree

Superpowers is ideal for developers who want AI agents to follow reliable engineering practices rather than generate code impulsively. Its philosophy emphasizes tests first, systematic debugging, reduced complexity, and evidence over claims.