一个由社区驱动的可重用AI代理工作流程注册库

什么是SkillHub？

SkillHub 是一个用于 AI 代理工作流程的包管理器，类似于 Homebrew 管理包或 npm 管理 JavaScript 库的方式。当 AI 编码代理（如 Claude Code、GitHub Copilot 或 ChatGPT）完成复杂任务时，它们会遵循特定的步骤序列。SkillHub 允许你：

• 存储已完成的AI代理工作流程作为可重复使用的“技能”
• **搜索并发现与任务相关的技能
• **在不同环境/硬件上提取并执行技能
• 通过GitHub拉取请求贡献新技能

问题

当AI代理完成复杂任务（如模型基准测试、开发环境搭建或部署应用）时，工作流程在完成后丢失。用户必须从头重新提示类似任务，浪费时间和精力。

解决方案

SkillHub将这些工作流程存储为结构化的markdown文件，包含：

• 逐步说明
• 代码示例
• 故障排除指南
• 成功标准
• 可搜索的元数据

快速启动

运用技能

贡献技能

SkillHub Registry

A community-driven registry of reusable AI agent workflows

What is SkillHub?

SkillHub is a package manager for AI agent workflows, similar to how Homebrew manages packages or npm manages JavaScript libraries. When AI coding agents (like Claude Code, GitHub Copilot, or ChatGPT) complete complex tasks, they follow specific sequences of steps. SkillHub allows you to:

• Store completed AI agent workflows as reusable “skills”
• Search and discover relevant skills for your tasks
• Pull and execute skills on different environments/hardware
• Contribute new skills via GitHub Pull Requests

The Problem

When AI agents complete complex tasks (like benchmarking a model, setting up a development environment, or deploying an application), the workflow is lost after completion. Users must re-prompt similar tasks from scratch, wasting time and effort.

The Solution

SkillHub stores these workflows as structured markdown files with:

• Step-by-step instructions
• Code examples
• Troubleshooting guides
• Success criteria
• Metadata for searchability

Quick Start

Using Skills

# Install the SkillHub CLI
pip install skillhub

# Search for skills
skillhub search "benchmark model"

# Pull a skill to your local directory
skillhub pull benchmark-qwen

# List all available skills
skillhub list

# View skill details
cat benchmark-qwen.md

Contributing Skills

Easy Way (Recommended): Submit via GitHub Issue - no setup required!

1. Go to Submit a Skill
2. Paste your skill markdown
3. Submit the issue
4. A Pull Request is created automatically!

Alternative: Manual PR submission

# Fork this repository
# Create a new skill using the template
cp templates/skill-template.md skills/category/my-skill.md

# Edit the skill file
# Validate your skill
python scripts/validate_skill.py skills/category/my-skill.md

# Submit a Pull Request

See CONTRIBUTING.md for detailed guidelines.

Available Skills

Benchmarking

• benchmark-qwen - Benchmark Qwen 3B model on custom hardware with comprehensive performance metrics

Data Engineering

• etl-pipeline - Set up a production-ready ETL pipeline with data validation and monitoring

Web Development

• react-app-setup - Set up a modern React application with TypeScript, Tailwind CSS, and best practices

Automation