obra/superpowers
An agentic skills framework & software development methodology that works.
Shell · ★ 319,506 · Fork 28,486 · 在 GitHub 查看
在手机上浏览源码、订阅每日 AI 趋势日报与项目架构深度解读:获取 CopoHub 客户端(iOS & HarmonyOS)→
GitHub Trending 上榜记录
obra/superpowers 已在 GitHub Trending 每日榜单上榜 132 天,最高排名第 1,首次上榜 2026-01-10,最近一次上榜 2026-11-18。
| 日期 | 排名 | Star | 当日增长 |
|---|---|---|---|
| 2026-11-18 | #9 | 319,506 | +1,917 |
| 2026-11-17 | #9 | 319,288 | +1,905 |
| 2026-11-01 | #8 | 310,288 | +665 |
| 2026-10-31 | #8 | 310,028 | +1,259 |
| 2026-10-30 | #2 | 309,390 | +1,255 |
| 2026-10-29 | #2 | 308,485 | +1,095 |
| 2026-10-28 | #3 | 307,401 | +1,096 |
| 2026-10-27 | #3 | 306,391 | +608 |
| 2026-10-17 | #8 | 300,915 | +973 |
| 2026-10-16 | #6 | 300,650 | +973 |
| 2026-10-15 | #6 | 299,842 | +688 |
| 2026-10-04 | #11 | 295,055 | +577 |
| 2026-10-03 | #3 | 294,883 | +578 |
| 2026-10-02 | #3 | 294,420 | +561 |
| 2026-10-01 | #7 | 293,927 | +594 |
| 2026-09-26 | #4 | 291,812 | +468 |
| 2026-09-25 | #4 | 291,627 | +611 |
| 2026-09-24 | #5 | 291,196 | +606 |
| 2026-09-23 | #5 | 290,648 | +528 |
| 2026-09-11 | #3 | 284,929 | +732 |
| 2026-09-10 | #3 | 284,666 | +731 |
| 2026-09-09 | #3 | 283,984 | +690 |
| 2026-09-08 | #7 | 283,316 | +446 |
| 2026-09-04 | #19 | 281,591 | +462 |
| 2026-09-03 | #19 | 281,296 | +470 |
| 2026-08-23 | #4 | 276,424 | +592 |
| 2026-08-22 | #4 | 276,148 | +790 |
| 2026-08-21 | #4 | 275,612 | +789 |
| 2026-08-20 | #4 | 274,893 | +749 |
| 2026-08-19 | #7 | 274,219 | +514 |
| 2026-08-08 | #5 | 268,984 | +782 |
| 2026-08-07 | #5 | 268,704 | +858 |
| 2026-08-06 | #8 | 268,034 | +931 |
| 2026-08-05 | #5 | 267,243 | +931 |
| 2026-08-04 | #5 | 266,410 | +777 |
| 2026-07-30 | #10 | 263,642 | +616 |
| 2026-07-29 | #10 | 263,228 | +686 |
| 2026-07-26 | #8 | 261,319 | +479 |
| 2026-07-25 | #8 | 261,041 | +600 |
| 2026-07-12 | #21 | 252,703 | +740 |
| 2026-07-11 | #7 | 252,402 | +1,013 |
| 2026-07-10 | #7 | 251,732 | +1,096 |
| 2026-07-09 | #9 | 250,367 | +1,116 |
| 2026-07-08 | #9 | 249,753 | +999 |
| 2026-07-04 | #15 | 245,803 | +1,209 |
| 2026-07-03 | #6 | 245,480 | +1,205 |
| 2026-07-02 | #6 | 244,327 | +1,047 |
| 2026-07-01 | #18 | 243,055 | +890 |
| 2026-06-30 | #16 | 242,421 | +884 |
| 2026-06-20 | #12 | 233,718 | +1,110 |
| 2026-06-19 | #4 | 233,278 | +1,429 |
| 2026-06-18 | #4 | 232,341 | +1,435 |
| 2026-06-17 | #5 | 230,972 | +1,205 |
| 2026-06-14 | #4 | 227,339 | +924 |
| 2026-06-13 | #4 | 226,865 | +1,275 |
| 2026-06-12 | #6 | 225,927 | +1,322 |
| 2026-06-11 | #7 | 224,739 | +1,323 |
| 2026-06-10 | #7 | 223,527 | +1,205 |
| 2026-06-07 | #12 | 219,982 | +700 |
| 2026-06-06 | #12 | 219,606 | +1,008 |
同期上榜的项目
- elementalsouls/Claude-BugHunter – A Claude Code skill bundle for bug hunting and external red-team work - 82 skills, 15 slash commands, 681 disclosed-report patterns curated across 24 core vulnerability classes, plus enterprise identity + infrastructure attack matrices.
- mvanhorn/last30days-skill – AI agent skill that researches any topic across Reddit, X, YouTube, HN, Polymarket, and the web - then synthesizes a grounded summary
- alibaba/open-code-review – Secure, fast, efficient, battle-tested at Alibaba's scale. Hybrid architecture code review tool: deterministic pipelines + LLM Agent, precise line-level comments, built-in multi-language ruleset (NPE, thread-safety, XSS, SQL injection), OpenAI & Anthropic compatible.
- Tencent/LoopForge – Self-hosted workflows for coding-agent delivery. Orchestrate CodeBuddy, Codex and Claude Code with scripts and human approval gates. Includes Local workflows for coding assistants.
- dream-num/univer – Build your own AI office with Univer Office SDK. Bring editing, calculation, collaboration, and file processing into your product for people and AI agents. Built on an extensible open-source core.
- getarcaneapp/arcane – Modern Docker Management, Designed for Everyone
obra/superpowers 项目深度剖析与商业评估报告
obra/superpowers 项目深度剖析与商业评估报告
1. 核心定位与商业落地价值
1.1 根本痛点 Superpowers 直击当前 AI 编码代理(Coding Agent)领域的核心痛点:LLM 的"低方差"输出质量问题。裸用 Claude Code、Gemini CLI 等工具时,模型在长任务中极易出现目标漂移、测试跳过、需求误解、上下文遗忘等行为。该项目通过将软件工程方法论(TDD、YAGNI、DRY)编码为可强制触发的 Agent 技能,将开发过程从"模型自由发挥"转变为"受控流水线",从根本上解决 AI 编码的不可预测性问题。
1.2 企业级商业落地场景 标准化 AI 研发流水线:企业可将 Superpowers 作为内部 AI 开发规范层,强制所有 Agent 遵循统一的 brainstorm → plan → TDD → review → merge 流程,确保产出代码质量可审计、可回溯。 大型遗留系统维护:其 systematic-debugging 技能(4 阶段根因分析)和 verification-before-completion 机制,非常适合处理复杂遗留代码库的缺陷修复,减少 AI 误改风险。 多 Agent 并行开发:subagent-driven-development 和 dispatching-parallel-agents 支持企业将大型重构任务分解为多个隔离 worktree 并行执行,显著提升交付速度。
1.3 竞品差异化与护城河 与 Cursor 的规则文件、GitHub Copilot 的指令集相比,Superpowers 的护城河在于: 跨平台兼容性:已适配 14+ 编码代理(Claude Code、Gemini CLI、Codex、Cursor、Devin 等),形成事实上的"Agent 技能层标准"。 方法论深度:不是简单的 prompt 集合,而是完整的 SDLC 流程封装,包含两阶段代码审查、红绿 TDD 强制、计划粒度控制(2-5 分钟任务)等工程级约束。 社区网络效应:27.6 万星标、2.4 万 fork,叠加 Prime Radiant 的商业支持([email protected]),形成开源社区 + 企业服务的双轮驱动。
2. 核心架构设计与技术深度
2.1 模块拆解思想 项目采用声明式技能(Skill)架构,每个技能是一个自包含的 SKILL.md 文件,包含触发条件、执行步骤、输入输出契约。这种设计遵循 Unix 哲学——"每个技能做一件事并做好",通过组合而非继承实现复杂流程: 触发器(Triggers):基于会话上下文自动激活(如检测到"设计已批准"即触发 writing-plans),实现零人工干预的流程编排。 状态机流转:核心链路严格遵循 brainstorming → worktrees → writing-plans → subagent-driven-development → TDD → code-review → finish-branch 的线性状态机,每个阶段有明确的进入/退出条件。
2.2 核心链路数据流 用户意图 → brainstorming(需求澄清,产出设计文档) → using-git-worktrees(隔离分支,验证测试基线) → writing-plans(任务分解,每个任务含精确文件路径+完整代码+验证步骤) → subagent-driven-development(每任务派发新 subagent,两阶段审查) → test-driven-development(RED-GREEN-REFACTOR 循环) → requesting-code-review(严重度分级,阻断性问题阻止推进) → finishing-a-development-branch(合并/PR/保留/丢弃决策)
在 README 中展示上榜徽章
[](https://github.fq6825.top/repo/obra/superpowers)