AG2 (AutoGen) — Multi-Agent Conversation Framework

You are an expert in AG2 (formerly AutoGen), the opensource multiagent conversation framework. You help developers build systems where multiple AI agents collaborate through structured conversations — with tool use, humanintheloop, code execution, group chat orchestration, and nested conversations — for complex tasks l

概览

AG2, previously known as AutoGen, is an open-source framework designed to facilitate complex workflows through multi-agent collaboration. According to the TerminalSkills/skills repository, this framework allows developers to orchestrate structured conversations between multiple AI agents. Key features include integrated tool usage, human-in-the-loop interactions, and autonomous code execution. The system supports various communication patterns, such as group chat orchestration and nested conversations, to tackle sophisticated tasks. By leveraging AG2, developers can create modular systems where specialized agents work together. The framework is compatible with major models like Gemini and Claude, providing a flexible foundation for building autonomous applications that require high-level reasoning and coordinated multi-step processes.

使用场景

Automating software development through collaborative agent coding and peer review.
Executing multi-step research and data analysis by delegating sub-tasks to specialized agents.
Creating interactive applications that integrate human feedback and external APIs within conversational flows.

安装说明

# Review source first
open https://github.com/TerminalSkills/skills/blob/main/skills/ag2/SKILL.md

Copy or clone the skill folder into your agent skills directory after reviewing its instructions and scripts.

安全提示

When implementing AG2, developers should carefully manage code execution environments and API access. Since the framework supports autonomous code execution and tool use, it is essential to define clear boundaries for agent permissions and monitor interactions to prevent unauthorized actions or data exposure during multi-agent sessions.

相关 Skills

Constant Time Testing

trailofbits/skills

前端设计

通过运行密码学实现来测量其中的时序侧信道,使用 dudect 进行统计分析,并使用基于 Valgrind 的 Timecop 进行动态跟踪。涵盖形式化、符号化、动态和统计类工具以及如何解读结果。适用于测试正在运行的实现是否为恒定时间、测量已编译二进制文件的时间方差,或调查疑似的时序攻击。不适用于静态检查编译器输出——constant-time-analysis 插件

Claude CodeClaude
securitytesting
7,190 Stars已链接来源

GSAP Core

mxyhi/ok-skills

前端设计

核心 API 的官方 GSAP skill —— 包含 gsap.to()、from()、fromTo()、easing、duration、stagger、defaults、gsap.matchMedia()(响应式、prefers-reduced-motion)。当用户请求 JavaScript 动画库、React/Vue/vanilla 中的动画、GSAP 补间动画、缓动、基础动画、响应式或减弱动态效果 a 时使用。

CodexClaude Code
reactsecurity
423 StarsApache-2.0

花叔Design · Huashu-Design

mxyhi/ok-skills

前端设计

花叔Design——用HTML做高保真原型、交互Demo、幻灯片、动画、设计变体探索+设计方向顾问+专家评审。根据任务embody对应专家(UX/动画师/幻灯片设计师/原型师),避免web design tropes。触发词:做原型、交互原型、HTML演示、动画Demo、设计变体、hi-fi设计、UI mockup、prototype、做个HTML页面、做个可视化、app原型、iOS原型、导出MP4/GIF、60fps视频、设计风格、设计方向、配色方案、推荐风格、选个风格、做个好看的、评审、好不好看、review this design、带解说的动画、解说视频、长视频科普、voiceover、narration、5分钟讲清楚什么是X

CodexClaude Code
pythonreact
423 StarsApache-2.0

Design Build

tinh2/skills-hub-registry

前端设计

快速构建独特、生产级的界面。交付可运行的代码(Web 或移动端),具备现代化的 CSS、流畅的排版、富有目的性的动效,且毫无平庸千篇一律的 AI 审美。适用场景:“build UI”、“design this”、“create interface”、“make it look good”、“build screen”、“implement design”、“frontend build”、“design build”。

CodexClaude Code
reactdesign
16 Stars已链接来源

Design Audit

tinh2/skills-hub-registry

前端设计

全面自主的设计质量审查,涵盖无障碍访问(WCAG 2.2)、性能、主题化、响应式/自适应设计以及反模式检测。触发词:'audit design'、'check accessibility'、'a11y audit'、'design review'、'UI quality check'。

CodexClaude Code
reactdesign
16 Stars已链接来源

Design Claude

tinh2/skills-hub-registry

前端设计

以 HTML 为先的高保真设计技能:交互式原型、幻灯片演示文稿、动画演示、设计方案探索、视觉方向建议与专业评审。触发词:制作原型、设计演示、交互式原型、HTML 演示文稿、动画演示。

CodexClaude Code
pythonreact
16 Stars已链接来源