Superpowers
Superpowers is an engineering discipline framework plugin for Claude Code, designed to regulate AI-assisted programming behavior through preset rules and constraint mechanisms. The plugin includes multiple built-in engineering skill modules and guides developers toward software engineering best practices such as Test-Driven Development (TDD) and design-first approaches through enforcement systems such as Iron Laws, targeting teams or individuals who need to maintain code quality and development standards in AI-assisted environments.
Core Facts
Timeline (last 90 days)
The tutorial covers a path starting from zero: installation, configuration, core Skills, then in order going through requirements clarification, solution design, task planning, development, testing, Review, and final acceptance.
The video's author has put together a Superpowers project hands-on toolkit, including the complete development workflow, a quick reference for core Skills, hands-on Prompts, and a project acceptance Checklist; per the video's instructions, it can be obtained by replying with a keyword in the comments.
Superpowers defines a phased software development process for coding agents, which in order includes requirements clarification, solution design, task breakdown, TDD development, independent review, and final acceptance.
Superpowers defines a phased software development process for coding agents such as Codex Cloud Code.
The Superpowers workflow includes: intercepting the prompt, enforcing Socratic clarification and outputting a design spec; after approval, using git worktree to create an isolated branch and verify a clean test baseline; breaking the spec into micro-tasks of 2 to 5 minutes; handing them to sub-agents for execution, where a sub-agent first deletes the code, forces out a failing test, then implements, and the sub-agent terminates after the task passes code review.
According to the video's data, with Superpowers the main session can stay under 15,000 tokens from start to PR throughout.
According to the video, Jesse Vincent's Superpowers has gained nearly 300,000 stars and is a complete development methodology that solves context pollution through sub-agent orchestration.
PStack has no dedicated maintenance skill responsible for "bringing problems back to the planning phase"; this can be filled in with maintenance-type skills such as Superpowers' SuperCollab.
To assess whether Superlite's "faster" claim holds, one must first clarify which specific Superpowers implementation it replaces; otherwise, the performance comparison lacks a baseline.
Superlite puts forward a clear value proposition: a faster alternative to Superpowers.
28 more timeline events
All Facts (20)
Superpowers abstracts the development process into a composable Skills system, including Brainstorming, Writing Code, Writing Tests, and Inspecting.
75%VerifiedSuperpowers is an open-source AI programming methodology framework.
70%UnverifiedSuperpowers的Writing Code技能遵循YAGNI和DRY（Don't Repeat Yourself）原则
95%UnverifiedSuperpowers的TDD流程要求AI先编写失败的测试用例，再编写刚好能让测试通过的最小实现代码
95%UnverifiedSuperpowers引入了TDD（测试驱动开发）和YAGNI（You Aren't Gonna Need It）两大软件工程原则
95%UnverifiedSuperpowers定义的指令优先级从高到低为：用户明确指令、Superpowers的Scale规则、Coding Agent自身的Mode和Assistant Prompt
95%UnverifiedSuperpowers的技术实现包含14个独立的Scale模块和一个Bootstrap Prompt
95%UnverifiedSuperpowers将任务拆解为五个关键阶段：需求确认、任务计划、任务执行、调试和代码审查
95%UnverifiedSuperpowers的TDD铁律只有三个例外：一次性原型代码、自动生成的代码和配置文件
90%UnverifiedSuperpowers与Cloud.md应配合使用：Cloud.md管控代码规范，Superpowers管控开发流程
90%UnverifiedSuperpowers的Skills可以根据项目需要自由组合，简单脚本可只用部分技能，生产级服务可启用全部技能
90%UnverifiedSuperpowers的子Agent完成工作后，主Agent会进行代码审查和检验，模拟真实团队中的Code Review流程
90%UnverifiedSuperpowers的核心理念是Spec First，即在编写代码之前通过结构化问题澄清需求和意图
90%Unverified一个功能从想法到上线需要经过七步流程：需求产物生成、Auto Plan评审、TDD编码、Review审查、QA验收、SHIP发布、Archive归档
85%UnverifiedSuperpowers通过Hard Gate机制自动拦截低质量代码，每段业务逻辑必须先写测试，测试跑不过就不能往下走
85%UnverifiedSuperpowers的模块化设计体现了软件工程中的单一职责原则（SRP），每个Scale专注于一个特定阶段的规则
85%Unverified不同LLM对Bootstrap Prompt的响应程度存在差异，这是Superpowers需要做兼容性适配的根本原因
85%UnverifiedSuperpowers会显著增加每次交互的Token消耗，通过注入Bootstrap Prompt和多个Scale模块会增加数百至数千个Token的固定开销
85%UnverifiedSuperpowers本质上是一个专为Coding场景设计的轻量级Agent Harness实现
85%Unverified在AI辅助开发中，计划阶段的错误会导致执行阶段反复试错，造成Token大量浪费，正确的计划可以将Token消耗减少约一半
70%