ai-agents-public
💡 摘要
一个全面的AI代理和Claude代码技能库,适用于各种应用。
🎯 适合人群
🤖 AI 吐槽: “看起来很能打,但别让配置把人劝退。”
风险:Medium。建议检查:是否执行 shell/命令行指令;是否发起外网请求(SSRF/数据外发);API Key/Token 的获取、存储与泄露风险;文件读写范围与路径穿越风险。以最小权限运行,并在生产环境启用前审计代码与依赖。
AI Agents Library
Production-Ready AI Agent Prompts & Claude Code Skills
A curated collection of 40+ Custom GPT agents and 50 Claude Code skills optimized for ChatGPT, Claude Projects, Codex, and Gemini platforms.
Quick Start • Custom GPTs • Claude Skills • Frameworks • Contributing
Table of Contents
- What's Inside
- Repository Architecture
- Custom GPT Agents
- Claude Code Skills
- Frameworks
- Quick Start
- Prompt Engineering Framework
- Platform Support
- Contributing
- License
What's Inside
This repository contains 40+ specialized AI agents and 50 Claude Code skills organized by domain, each optimized for specific tasks and workflows. All agents follow a consistent template structure with platform-specific optimizations.
graph TB subgraph "AI Agents Library" A[Custom GPT Agents<br/>40+ Specialized Agents] --> A1[Education] A --> A2[Lifestyle] A --> A3[Productivity] A --> A4[Programming] A --> A5[Research & Analysis] A --> A6[Writing] B[Claude Code Skills<br/>50 Development Skills] --> B1[Software Development] B --> B2[AI/ML Engineering] B --> B3[DevOps & Platform] B --> B4[Quality & Testing] B --> B5[Documentation] C[Frameworks<br/>Copy-Paste Ready] --> C1[Claude Code Kit] C --> C2[Codex Kit] C --> C3[Gemini Kit] D[Master Template v3.5<br/>13-Section Framework] --> A D --> B D --> C end style A fill:#4A90E2 style B fill:#9B59B6 style C fill:#E67E22 style D fill:#27AE60
What Makes This Different?
- Production-Ready: Every agent tested and optimized for real-world use
- Multi-Platform: Works across ChatGPT, Claude, Codex, and Gemini
- Systematic Framework: Master Template v3.5 with 13 standardized sections
- Character-Optimized: Custom GPT agents under 8000 char limit
- Curated Resources: JSON source files with 60+ curated references per domain
- Copy-Paste Kits: Complete
.claude/and.codex/workspace templates
Repository Architecture
graph LR subgraph "Repository Structure" A[custom-gpt/] --> A1[education/] A --> A2[lifestyle/] A --> A3[productivity/] A --> A4[programming/] A --> A5[research-n-analysis/] A --> A6[writing/] B[frameworks/] --> B1[claude-code-kit/<br/>50 Skills] B --> B2[codex-kit/<br/>Router System] B --> B3[gemini-kit/<br/>Router System] C[Each Agent] --> C1[01_agent-name.md<br/>Main Prompt] C --> C2[02_sources.json<br/>Curated Resources] C --> C3[agent-name.yaml<br/>Configuration] end style A fill:#4A90E2,color:#fff style B fill:#9B59B6,color:#fff style C fill:#27AE60,color:#fff
Three-File Agent Pattern
Every agent follows a consistent structure:
AgentName/ ├── 01_agent-name.md # Main prompt file (<8000 chars for Custom GPT) ├── 02_sources-agent-name.json # Curated web resources and references ├── agent-name.yaml # Configuration (role, commands, constraints) └── archive/ # Version history (git-ignored)
Custom GPT Agents
40+ specialized agents optimized for ChatGPT Custom GPTs (8000 character limit).
Agent Categories
Education (3 agents)
- English Tutor - IELTS Writing Task 1 & 2 preparation with band score feedback
- UK Tax and Legal Adviser 2025 - UK tax, legal, and regulatory guidance (2025/26 tax year)
- Life in the UK 2025 - Citizenship test preparation with practice questions
Lifestyle (7 agents)
- Fitness Buddy - Personalized workout programming with progressive overload strategies
- DietGPT - Comprehensive nutrition planning with macro/micronutrient targets
- Sleep Coach - Sleep optimization with CBT-I techniques and environmental guidance
- ChildBridge - Child development expert with parenting strategies
- Pet Whisperer - Pet care covering nutrition, training, and species-specific needs
- ReelRecipe - Short-form video content strategist for TikTok, Instagram Reels, YouTube Shorts
- CineMatch - Movie recommendations with 66+ sources and 17 discovery commands
Productivity (5 agents)
- Prompt Engineer - Master template v3.5 with 4 deployment modes and platform-specific versions
- Product Coach - Product management with discovery, roadmapping, and prioritization frameworks
- The Negotiator - Principled negotiation with BATNA analysis and tactical empathy
- Contract Crusher - Legal contract analysis with risk identification and plain-English summaries
- SMMA - Social media marketing with campaign strategy and paid advertising tactics
Programming (5 agents)
- AI Agents Builder - OpenAI Assistants API, LangChain, CrewAI with multi-agent orchestration
- LLM Engineer - Fine-tuning, RAG architecture, embeddings, and production deployment
- Data Scientist - ML/AI with PyTorch, TensorFlow, XGBoost, and MLOps pipelines
- PRD Business Analyst - User stories, technical specifications, and API contracts
- SQL and DevOps Engineer - Database optimization, Kubernetes, Terraform, and observability
Research & Analysis (3 agents)
- AI Strategist - AI readiness assessments, use case identification, and governance frameworks
- Startup Consultant - Business model validation, fundraising strategy, and financial modeling
- Strategy Consultant - Porter's Five Forces, SWOT, PESTEL, and competitive positioning
Writing (2 agents)
- AI Text Humaniser - Transform AI content into natural writing while eliminating detection markers
- FAANG Resume Coach - ATS optimization, STAR stories, and behavioral interview preparation
Claude Code Skills
50 production-ready skills for Claude Code, organized into domain-specific categories.
graph LR subgraph "Claude Code Skills (50)" A[Software Development] --> A1[Frontend<br/>Backend<br/>Architecture] B[AI/ML Engineering] --> B1[LLM Engineering<br/>Agents<br/>Data Science] C[DevOps & Platform] --> C1[Database<br/>Platform<br/>Document Automation] D[Quality & Testing] --> D1[Code Review<br/>Testing<br/>Debugging<br/>Refactoring] E[Documentation] --> E1[API Design<br/>Codebase Audit<br/>Technical Writing] end style A fill:#3498DB style B fill:#9B59B6 style C fill:#E67E22 style D fill:#2ECC71 style E fill:#34495E,color:#fff
Installation
Copy the entire .claude/skills/ directory to your Claude Code workspace:
# Clone the repository git clone https://github.com/vasilyu1983/AI-Agents-public cd AI-Agents-public # Copy skills to Claude Code cp -r frameworks/claude-code-kit/initial-setup/skills ~/.config/claude-code/skills/
Frameworks
Three complete development kits for different AI platforms:
Claude Code Kit
Complete .claude/ workspace with 50 skills covering software development, AI/ML, DevOps, and quality engineering.
Copy-paste ready: Drop the initial-setup/ directory into your Claude Code workspace.
Codex Kit
Router-based system for Codex CLI with intelligent agent dispatch and workflow orchestration.
Gemini Kit
Gemini-optimized router system with platform-specific adaptations.
See frameworks documentation →
Quick Start
Using a Custom GPT Agent
- Browse the agent catalog to find an agent
- Navigate to the agent folder (e.g.,
custom-gpt/productivity/Prompt Engineer/) - Copy content from
01_agent-name.md(guaranteed <8000 chars) - Create new Custom GPT in ChatGPT
- Paste into Instructions field
- Upload
02_sources-*.jsonas knowledge file (optional)
Installing Claude Code Skills
# Clone repository git clone https://github.com/vasilyu1983/AI-Agents-public # Install all skills cp -r frameworks/claude-code-kit/initial-setup/skills ~/.config/claude-code/skills/ # Verify installation ls ~/.config/claude-code/skills/
Using with Claude Projects
- Create new Claude Project
- Add
01_agent-name.mdas project knowledge - Add supplemental files (
02_sources-*.json) if available - Reference in custom instructions
Prompt Engineering Framework
All agents built on Master Template v3.5 with 13 standardized sections:
- VARS - Configuration variables
- IDENTITY - Role definition
- CONTEXT - Background knowledge
- CONSTRAINTS - Boundaries and requirements
- PRECEDENCE & SAFETY - Security protocols
- OUTPUT CONTRACT - Format specifications
- FRAMEWORKS - Reasoning frameworks
- WORKFLOW - Execution process
- ERROR RECOVERY - Edge case handling
- TOOLS & UI - Tool usage guidelines
- MEMORY - State management
- COMMANDS - Slash commands
- EXEMPLARS - Real-world examples
Template Features
graph TB A[Master Template v3.5] --> B[4 Deployment Modes]
优点
- 多种专业代理和技能的广泛选择
- 为实际使用优化的生产就绪状态
- 多平台支持提高可用性
- 结构化模板便于轻松定制
缺点
- 复杂性可能会让新用户感到困惑
- 对高级定制的指导有限
- 依赖外部平台以实现全部功能
相关技能
免责声明:本内容来源于 GitHub 开源项目,仅供展示和评分分析使用。
版权归原作者所有 vasilyu1983.
