addyosmani/agent-skills
Production-grade engineering skills for AI coding agents.
JavaScript · ★ 118,747 · Fork 12,379 · 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
addyosmani/agent-skills 已在 GitHub Trending 每日榜单上榜 50 天,最高排名第 1,首次上榜 2026-05-06,最近一次上榜 2026-11-15。
| 日期 | 排名 | Star | 当日增长 |
|---|---|---|---|
| 2026-11-15 | #41 | 117,745 | +411 |
| 2026-11-14 | #14 | 117,554 | +591 |
| 2026-11-13 | #9 | 117,126 | +816 |
| 2026-11-12 | #5 | 116,630 | +812 |
| 2026-11-11 | #3 | 115,892 | +788 |
| 2026-11-10 | #3 | 115,110 | +465 |
| 2026-10-10 | #8 | 104,271 | +436 |
| 2026-10-09 | #8 | 103,933 | +751 |
| 2026-10-08 | #6 | 103,139 | +677 |
| 2026-10-07 | #6 | 102,733 | +693 |
| 2026-10-05 | #12 | 101,360 | +336 |
| 2026-10-04 | #10 | 101,185 | +336 |
| 2026-10-03 | #10 | 100,810 | +305 |
| 2026-09-21 | #13 | 97,940 | +736 |
| 2026-09-20 | #3 | 97,634 | +729 |
| 2026-09-19 | #3 | 96,972 | +675 |
| 2026-09-18 | #3 | 96,357 | +680 |
| 2026-09-17 | #3 | 96,005 | +680 |
| 2026-09-16 | #12 | 95,373 | +656 |
| 2026-09-15 | #12 | 94,710 | +386 |
| 2026-09-04 | #12 | 92,163 | +264 |
| 2026-09-03 | #12 | 92,003 | +280 |
| 2026-08-30 | #18 | 90,845 | +196 |
| 2026-08-29 | #18 | 90,691 | +188 |
| 2026-08-12 | #4 | 86,423 | +578 |
| 2026-08-11 | #4 | 86,186 | +659 |
| 2026-08-10 | #4 | 85,703 | +680 |
| 2026-08-09 | #2 | 85,070 | +779 |
| 2026-08-08 | #2 | 84,483 | +1,131 |
| 2026-08-07 | #2 | 83,823 | +1,131 |
| 2026-08-06 | #2 | 82,841 | +588 |
| 2026-08-05 | #6 | 81,945 | +203 |
| 2026-07-11 | #4 | 77,078 | +1,116 |
| 2026-07-10 | #2 | 76,759 | +2,554 |
| 2026-07-09 | #1 | 75,781 | +2,582 |
| 2026-07-08 | #1 | 73,850 | +1,322 |
| 2026-07-07 | #2 | 72,037 | +1,311 |
| 2026-07-06 | #2 | 70,711 | +1,114 |
| 2026-06-14 | #2 | 58,906 | +1,514 |
| 2026-06-13 | #1 | 58,243 | +2,656 |
| 2026-06-12 | #1 | 56,687 | +3,278 |
| 2026-06-11 | #1 | 54,531 | +3,275 |
| 2026-06-10 | #1 | 51,558 | +781 |
| 2026-06-09 | #16 | 49,768 | +348 |
| 2026-05-11 | #3 | 39,093 | +1,065 |
| 2026-05-10 | #3 | 38,313 | +3,009 |
| 2026-05-09 | #2 | 37,310 | +2,801 |
| 2026-05-08 | #2 | 35,185 | +3,062 |
| 2026-05-07 | #2 | 32,808 | +3,058 |
| 2026-05-06 | #2 | 30,270 | +629 |
同期上榜的项目
- 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.
- ag-ui-protocol/ag-ui – AG-UI: the Agent-User Interaction Protocol. Bring Agents into Frontend Applications.
- google-labs-code/stitch-skills – A library of Agent Skills designed to work with the Stitch MCP server. Each skill follows the Agent Skills open standard, for compatibility with coding agents such as Antigravity, Gemini CLI, Claude Code, Cursor.
- backnotprop/plannotator – The workspace to stay in the loop and on top of agents - shape, decide, plan, annotate, review code, share with team, and send all feedback to agents with one click.
- ChromeDevTools/chrome-devtools-mcp – Chrome DevTools for coding agents
- 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.
agent-skills 项目深度剖析与商业评估报告
agent-skills 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
根本痛点:当前AI编码代理(Claude Code、Cursor、Gemini CLI等)默认行为倾向于“最短路径”——跳过规范、测试、安全审查和代码审查等关键工程步骤,产出质量不稳定、难以投入生产的代码。这导致企业采用AI编码时面临“原型可用、生产不可用”的困境。
商业落地场景: 企业AI编码标准化:将agent-skills作为企业AI编码的“标准操作规程”(SOP),确保所有AI生成的代码遵循统一的质量门禁(quality gates)。例如,金融科技公司可强制所有AI提交的代码必须通过/security-and-hardening和/code-review-and-quality技能。 CI/CD流水线集成:将技能作为预检步骤嵌入CI/CD管道,实现“AI代码自动审查-测试-安全扫描”的自动化流水线。可显著减少人工审查工作量,预计可降低30-50%的代码审查时间。 培训与知识管理:技能本身编码了Google工程文化中的最佳实践(Hyrum's Law、Beyonce Rule、Chesterton's Fence等),可作为新工程师的“数字导师”,加速团队工程文化统一。
竞品差异化与护城河: 结构化深度:不同于通用的prompt模板(如GitHub Copilot的指令文件),agent-skills提供了22个具有严格步骤、验证门禁和反合理化的完整工作流。每个技能包含Process → Rationalizations → Red Flags → Verification的完整结构,这是其他方案缺乏的工程级严谨性。 跨平台兼容性:支持Claude Code、Cursor、Gemini CLI、Windsurf、OpenCode、GitHub Copilot、Kiro IDE等7+主流AI编码工具,且采用纯Markdown格式,理论上可适配任何支持系统提示的代理。这种“一次编写、随处运行”的能力是核心护城河。 反合理化机制:独创的“反合理化表”(Anti-Rationalization Table)——每个技能列出代理常用的跳过步骤借口(如“我会稍后添加测试”)并内置反驳论据。这直接解决了AI代理“偷懒”的根本问题,是工程化落地的关键创新。
2. 核心架构设计与技术深度
模块拆解思想:采用生命周期驱动的模块化架构,严格遵循DEFINE → PLAN → BUILD → VERIFY → REVIEW → SHIP六阶段模型。每个阶段对应一组技能,技能之间通过using-agent-skills元技能实现路由分发。这种设计借鉴了Google的“Engineering Pipeline”理念,将软件开发视为可重复、可度量的流水线。
关键设计模式: 1. 代理-技能模式(Agent-Skill Pattern):agents/目录下的角色(code-reviewer、test-engineer、security-auditor)作为“代理”,调用skills/下的技能作为“能力”。这种分离使角色可复用技能(如code-reviewer可复用code-review-and-quality和security-and-hardening),技能也可被不同角色调用。 2. 渐进式披露(Progressive Disclosure):每个技能以SKILL.md为入口,仅加载核心流程;references/目录下的补充检查清单(testing-patterns.md、security-checklist.md等)仅在技能执行过程中按需引用。这直接针对AI代理的token预算限制——最小化上下文窗口占用,同时确保需要时能获取完整信息。 3. 命令-技能映射:7个斜杠命令(/spec、/plan、/build等)作为用户接口,每个命令映射到一组技能。例如/build触发incremental-implementation、test-driven-development、api-and-interface-design等多个技能的组合执行。这种设计降低了用户学习成本,同时保持了技能的可组合性。
在 README 中展示上榜徽章
[](https://github.fq6825.top/repo/addyosmani/agent-skills)