agents by 0xkynz/codekit
npx skills add https://github.com/0xkynz/codekit --skill agentsAGENTS.md 是一个用于指导 AI 编码代理的开放标准。它兼容 25 个以上的代理,包括 Claude Code、GitHub Copilot、Cursor、Devin、Windsurf、Cline、Aider 等。
将 AGENTS.md 作为适用于任何代理的通用基准。对于特定平台独有的功能,请使用平台特定文件:
| 文件 | 适用范围 | 最佳用途 |
|---|---|---|
AGENTS.md | 通用 —— 所有 AI 代理 | 项目概述、构建命令、代码风格、架构、测试 |
CLAUDE.md | 仅限 Claude Code | Claude 特定功能、记忆库、技能引用 |
.cursorrules | 仅限 Cursor | Cursor 特定行为 |
AGENTS.md is an open standard for guiding AI coding agents. It works across 25+ agents including Claude Code, GitHub Copilot, Cursor, Devin, Windsurf, Cline, Aider, and more.
Use AGENTS.md as the universal baseline that works with any agent. Use platform-specific files for features unique to that platform:
| File | Scope | Best For |
|---|---|---|
AGENTS.md | Universal — all AI agents | Project overview, build commands, code style, architecture, testing |
CLAUDE.md | Claude Code only | Claude-specific features, memory-bank, skill references |
广告位招租
在这里展示您的产品或服务
触达数万 AI 开发者,精准高效
GEMINI.md| 仅限 Gemini CLI |
| Gemini 特定上下文 |
经验法则: 如果某个指令适用于处理你项目的任何 AI 代理,就将其放入 AGENTS.md。如果是平台特定的,就放入对应的平台文件。
AGENTS.md 是标准的 Markdown 文件。没有必需的架构 —— 使用对你的项目有意义的标题。推荐的部分如下:
# 项目名称
## 项目概述
[项目功能、目的]
## 构建与测试命令
[如何安装、运行、测试、构建]
## 技术栈
[语言、框架、运行时、包管理器]
## 代码风格
[约定:TypeScript、分号、引号、格式化]
## 架构
[模式、目录结构、关键设计决策]
## 测试
[测试框架、覆盖率要求、如何运行测试]
## 依赖项
[关键包及其用途]
## 指导原则
[AI 代理规则:该做什么,避免什么]
在单体仓库中,在每个包/应用层级放置 AGENTS.md。距离被编辑代码最近的配置文件具有优先权:
monorepo/
├── AGENTS.md # 根级默认配置
├── packages/
│ ├── api/
│ │ └── AGENTS.md # API 特定规则(覆盖根级)
│ ├── web/
│ │ └── AGENTS.md # Web 应用规则(覆盖根级)
│ └── shared/
│ └── AGENTS.md # 共享库规则(覆盖根级)
codekit 可自动生成 AGENTS.md 以及其他平台文件:
# 为包括 AGENTS.md 在内的所有平台生成
codekit learn --all
# 仅生成 AGENTS.md
codekit learn -p agents
# 同步所有平台文件
codekit sync
生成的 AGENTS.md 包含:
codekit sync。AGENTS.md ← 通用基准(所有代理都读取此文件)
↕ 互补
CLAUDE.md ← Claude 特定(技能、记忆库指令)
.cursorrules ← Cursor 特定(IDE 行为)
GEMINI.md ← Gemini 特定(CLI 上下文)
memory-bank/ ← 会话连续性(所有文件均可引用)
所有这些文件都由 codekit learn 和 codekit sync 生成和同步。AGENTS.md 充当共同基础,确保任何 AI 代理 —— 无论是当前的还是未来的 —— 都能理解你的项目。
每周安装量
1
代码仓库
GitHub 星标数
1
首次出现
1 天前
安全审计
安装于
zencoder1
amp1
cline1
openclaw1
opencode1
cursor1
.cursorrules |
| Cursor only |
| Cursor-specific behaviors |
GEMINI.md | Gemini CLI only | Gemini-specific context |
Rule of thumb: If an instruction applies to any AI agent working on your project, put it in AGENTS.md. If it's platform-specific, put it in the platform file.
AGENTS.md is standard Markdown. No required schema — use headings that make sense for your project. Recommended sections:
# Project Name
## Project Overview
[What this project does, its purpose]
## Build and Test Commands
[How to install, run, test, build]
## Tech Stack
[Languages, frameworks, runtime, package manager]
## Code Style
[Conventions: TypeScript, semicolons, quotes, formatting]
## Architecture
[Patterns, directory structure, key design decisions]
## Testing
[Test framework, coverage requirements, how to run tests]
## Dependencies
[Key packages and their purposes]
## Guidelines
[Rules for AI agents: what to do, what to avoid]
In monorepos, place AGENTS.md at each package/app level. The nearest file to the code being edited takes precedence:
monorepo/
├── AGENTS.md # Root-level defaults
├── packages/
│ ├── api/
│ │ └── AGENTS.md # API-specific rules (overrides root)
│ ├── web/
│ │ └── AGENTS.md # Web app rules (overrides root)
│ └── shared/
│ └── AGENTS.md # Shared lib rules (overrides root)
codekit auto-generates AGENTS.md alongside other platform files:
# Generate for all platforms including AGENTS.md
codekit learn --all
# Generate only AGENTS.md
codekit learn -p agents
# Sync all platform files
codekit sync
The generated AGENTS.md includes:
codekit sync after major changes.AGENTS.md ← Universal baseline (all agents read this)
↕ complements
CLAUDE.md ← Claude-specific (skills, memory-bank instructions)
.cursorrules ← Cursor-specific (IDE behaviors)
GEMINI.md ← Gemini-specific (CLI context)
memory-bank/ ← Session continuity (referenced by all files)
All these files are generated and synced by codekit learn and codekit sync. AGENTS.md acts as the common denominator that ensures any AI agent — current or future — can understand your project.
Weekly Installs
1
Repository
GitHub Stars
1
First Seen
1 day ago
Security Audits
Installed on
zencoder1
amp1
cline1
openclaw1
opencode1
cursor1
AI新闻播客制作技能:实时新闻转对话式播客脚本与音频生成
1,200 周安装