What Is Superpowers? The Agent Skills Framework That Turns Claude Code into a Senior Developer
Introduction
AI coding agents can generate code at remarkable speed. But speed without discipline leads to untested implementations, context drift, and technical debt. Superpowers is an open-source agent skills framework designed to solve exactly this problem.
Created by Jesse Vincent (GitHub: obra) in October 2025, Superpowers transforms Claude Code from a "code generator" into a "disciplined senior developer." Accepted into the Claude Code official marketplace, the project has grown rapidly to become one of the most adopted tools in AI-assisted development.
Key Takeaways
- Superpowers applies a 7-phase development workflow to AI coding agents
- TDD (Test-Driven Development) is strongly enforced—code written without tests gets deleted
- Compatible with Claude Code, Cursor, Codex, OpenCode, GitHub Copilot CLI, and Gemini CLI
- Understanding how it compares to GStack, BMAD, and GSD is crucial for choosing the right tool
Why Has Superpowers Gained So Much Traction?
The core challenge with AI coding agents is that they produce code that is fast but unreliable. Superpowers addresses this through a unique lens: psychological persuasion principles applied to LLMs.
Research has validated that persuasion principles from Robert Cialdini's Influence: The Psychology of Persuasion are effective when applied to large language models. Superpowers leverages this insight through what Jesse Vincent calls the "1% Rule": if there is even a 1% chance a skill applies, the agent must invoke it. This prevents best practices from being skipped.
Traditional approaches relied on prompts to "please write tests" or "follow best practices"—requests that agents frequently ignored. Superpowers does not ask. It applies best practices through the structure of the workflow itself, with relevant skills triggering automatically based on context.
How Does the 7-Phase Workflow Operate?
The core of Superpowers is a composable development workflow consisting of seven phases. Each phase is implemented as an independent "skill" that activates automatically based on context.
1. Brainstorm
Before any code is written, the agent engages in Socratic questioning. It asks what you are actually trying to achieve and explores alternatives.
2. Spec
Requirements are clarified and formalized into a specification document. Ambiguous instructions are transformed into structured requirements.
3. Plan
The project is broken down into small tasks, each completable in 2–5 minutes. Independent tasks prevent context drift during long sessions.
4. TDD (Test-Driven Development)
The RED → GREEN → REFACTOR cycle is strictly enforced. If code is written without a corresponding failing test, Superpowers deletes it. This is the framework's most distinctive feature.
5. Subagent Development
Fresh subagents are spawned per task for parallel development. Each subagent operates with an independent context, minimizing drift during multi-hour autonomous sessions.
6. Review
Generated code undergoes systematic review for test coverage, code quality, and security concerns.
7. Finalize
Changes are consolidated, commit messages are generated, and documentation is updated.
How Do You Install and Use Superpowers?
Installation takes a single command. For Claude Code via the official marketplace:
/plugin install superpowers@claude-plugins-officialSuperpowers also supports Cursor, Codex, OpenCode, GitHub Copilot CLI, and Gemini CLI. For Codex and OpenCode, follow the INSTALL.md instructions in the official repository.
No configuration is required after installation. When you start working, the appropriate skills activate automatically based on your prompts. Say "I want to implement user authentication" and the Brainstorm skill kicks in. Say "write tests for this module" and the TDD skill takes over.
# Example: Development with Superpowers in Claude Code
claude "I want to implement user authentication"
# → Brainstorm → Spec → Plan → TDD proceeds automaticallyAccording to Jesse Vincent, Claude Code with Superpowers can work autonomously for hours at a time without deviating from the initial plan.
How Does Superpowers Compare to Other Frameworks?
As of 2026, the AI coding agent framework landscape is crowded. Here is how the major players compare.
| Framework | Popularity | Approach | Best For |
|---|---|---|---|
| Superpowers | Very high | Process constraints | TDD, solo/small teams |
| GStack | High | Decision constraints | Stack definition, context management |
| GSD | Growing | Execution constraints | Stability, parallel execution |
| BMAD | — | Agile lifecycle | Enterprise, team development |
| Hermes | — | Self-improving memory | Long-term projects, learning |
In one line: GStack thinks, GSD stabilizes, Superpowers executes.
- Superpowers constrains the development process. Write tests before code. Plan before you build.
- GStack constrains decision-making. It explicitly defines the tech stack so the agent works with the right context.
- GSD constrains the execution environment. It prioritizes stability through context isolation.
These frameworks are not mutually exclusive and can be combined. However, running multiple frameworks simultaneously can cause configuration conflicts, so choosing the right primary framework for your project is important.
The History and Future of Superpowers
The trajectory of Superpowers mirrors the evolution of AI coding agents themselves.
- October 2025: Jesse Vincent publishes "Superpowers: How I'm using coding agents," sharing effective development practices with Claude Code
- Early 2026: Official release as a Claude Code plugin. Accepted into the official marketplace and rapidly gains traction
- April 2026: Supports Claude Code, Cursor, Codex, OpenCode, GitHub Copilot CLI, and Gemini CLI, establishing itself as one of the most popular AI development tools on GitHub
As an MIT-licensed, fully open-source project, Superpowers has no usage restrictions for personal or commercial use. Community-driven development continues actively, with new skills and workflow improvements being contributed regularly.
FAQ
Q. Is Superpowers free to use?
A. Yes. It is an MIT-licensed open-source project. There are no subscriptions, usage limits, or hidden costs for personal or commercial use.
Q. Does it work with agents other than Claude Code?
A. Yes. It supports Cursor, Codex, OpenCode, GitHub Copilot CLI, and Gemini CLI. Refer to each platform's documentation for installation instructions.
Q. What if TDD enforcement feels too rigid?
A. Superpowers skills are composable by design. You can disable specific skills or customize them to fit your project's needs.
Q. Can I use it alongside GStack or GSD?
A. Technically yes, but configurations may conflict. The recommended approach is to choose one primary framework and selectively add complementary tools from others.
Q. Is there a steep learning curve?
A. No. Installation takes 30 seconds with zero configuration. Skills activate automatically based on context, so there are no special commands or workflows to memorize.
Summary
Superpowers tackles the fundamental challenge of AI coding agents—fast but sloppy—by enforcing structured workflows and applying psychological persuasion principles. Its 7-phase development pipeline, strict TDD enforcement, and composable skill architecture dramatically improve the quality of AI-generated code.
The rapid growth demonstrates that developers worldwide share this frustration. If you are working with AI coding agents, start by installing Superpowers from the official marketplace.
At ZenChAIne, we continue to cover AI agent development topics including what agent skills are and spec-driven development with agent skills.