Give Your AI Agent Superpowers — Stop Spaghetti Code with a Senior-Engineer Workflow

Ever get frustrated when your AI coding agent (like Claude or Cursor) takes your prompt and immediately starts writing a tangled mess of code without thinking it through?

https://github.com/obra/superpowers

Superpowers (obra/superpowers) isn’t just a prompt library—it’s a complete Agentic Skills Framework & Software Development Methodology that forces your AI to work like a disciplined Senior Engineer.

Here is everything you need to know to transform how your AI builds software.

🧠 The Philosophy

The core principles of Superpowers are mandatory workflows, not mere suggestions:

• Systematic over ad-hoc: Process over guessing.

• Test-Driven Development (TDD): Write tests first, always. It enforces the RED-GREEN-REFACTOR cycle. It even deletes code written before a test exists.

• Complexity reduction: Simplicity as the primary goal. Emphasizes YAGNI (You Aren't Gonna Need It) and DRY.

• Evidence over claims: Verify before declaring success.

⚙️ The Basic Workflow

Superpowers takes over from the moment you fire up your agent:

1. Brainstorming (Spec First): It stops the AI from jumping into code. It asks Socratic questions to tease out a spec, showing you the design in short, digestible chunks for sign-off.

2. Writing Plans: Once approved, the agent builds a crystal-clear implementation plan broken into 2-5 minute tasks. Each task gets exact file paths, expected code, and verification steps.

3. Execution & TDD: After you say "go", it launches into action. It creates an isolated Git worktree on a new branch, runs project setup, and verifies a clean test baseline.

4. Subagent-Driven Development: It dispatches a fresh subagent per task with a two-stage review process (spec compliance, then code quality). The AI can work autonomously for hours without deviating from your plan.

5. Finishing Up: When tasks complete, it verifies tests, presents options to merge/PR/discard, and cleans up the worktree.

🧰 What’s Inside? (The Core Skills)

Because these skills trigger automatically based on your conversation, you don't need to do anything special. Just ask for help, and the agent invokes the right superpower:

• brainstorming: Socratic design refinement before writing code.

• using-git-worktrees: Parallel development branches to keep your main workspace clean.

• writing-plans: Detailed, step-by-step implementation plans.

• subagent-driven-development / executing-plans: Dispatches subagents or executes in batches with human checkpoints.

• test-driven-development: The rigid RED-GREEN-REFACTOR cycle.

• requesting-code-review: A pre-review checklist that blocks progress if critical issues are found.

• systematic-debugging: A 4-phase root cause process to fix bugs properly.

🚀 Installation Guide

Installation varies depending on your AI coding platform:

Claude Code: Available via the official marketplace.

/plugin install superpowers@claude-plugins-official

(Alternatively, add the custom marketplace first: /plugin marketplace add obra/superpowers-marketplace then install superpowers@superpowers-marketplace)

Cursor: In the Cursor Agent chat, type:

/add-plugin superpowers

(Or search for "superpowers" in the plugin marketplace).

OpenCode / Codex: Tell your agent:

"Fetch and follow instructions from https://raw.githubusercontent.com/obra/superpowers/refs/heads/main/.opencode/INSTALL.md" (swap .opencode with .codex for Codex).

Gemini Extensions:

gemini extensions install https://github.com/obra/superpowers