AI Workflow & Prompt Library 🤖

A curated collection of specialized Claude subagents and production-ready prompts for AI-powered workflows across development, content creation, security, and product management.

What This Repository Is

This is a living knowledge base that transforms how you work with Claude Code and other AI systems. Instead of starting from scratch each time, you get:

• 20+ Specialized Subagents that act as domain experts
• Battle-tested Prompts for complex workflows
• CLAUDE.md Integration for seamless AI assistance in your projects
• Production-ready Templates you can immediately use

Think of it as your AI workforce - each subagent is a specialist that knows exactly how to help with specific tasks, from code reviews that prevent outages to SEO content that ranks.

Repository Structure

ai/
├── claude subagents/           # 20+ specialized AI agents
│   ├── code-reviewer.md       # Security-focused code review
│   ├── prompt-engineer.md     # LLM optimization expert
│   ├── backend-architect.md   # API & system design
│   ├── security-auditor.md    # OWASP compliance
│   └── seo-*.md              # Complete SEO workflow
├── prompt library/            # Production prompts & templates
│   ├── PM Copilot.md         # Product management coaching
│   └── System Prompt - *.md  # Specialized system prompts
├── CLAUDE.md                  # Repository guidance for Claude Code
└── README.md                  # This file

🎯 Understanding CLAUDE.md Files

CLAUDE.md files are game-changers for AI-assisted development. They're instruction files that tell Claude Code exactly how to work with your specific codebase.

Why CLAUDE.md Matters

Without CLAUDE.md:

User: "Run the tests"
Claude: "I don't know your test command. Is it npm test? pytest? cargo test?"

With CLAUDE.md:

User: "Run the tests"  
Claude: *immediately runs `npm run test:watch` because it knows your setup*

What Goes in CLAUDE.md

1. Build & Development Commands

   # Development Commands
   - Build: `npm run build`
   - Tests: `npm run test:watch`
   - Lint: `npm run lint:fix`

2. Architecture Overview

   # Architecture
   - `/src/components` - React components with TypeScript
   - `/src/hooks` - Custom hooks following useX pattern
   - `/src/utils` - Pure functions, no React dependencies

3. Project-Specific Context

   # Important Notes
   - Always run security audit after auth changes
   - Database migrations require staging deployment first
   - API rate limits: 100 req/min in development

CLAUDE.md Best Practices

• ✅ Include actual commands, not generic descriptions
• ✅ Explain non-obvious architecture decisions
• ✅ Document common gotchas and debugging steps
• ❌ Don't repeat what's obvious from file structure
• ❌ Don't include generic development advice

🤖 The Subagent System

Subagents are specialized AI personas designed for specific domains. Each has deep expertise and follows proven patterns.

How Subagents Work

Each subagent is a markdown file with:

---
name: code-reviewer
description: Expert code review specialist. Use immediately after writing code.
model: sonnet
---

# Specialized instructions and expertise...

Key Subagents Explained

🔍 code-reviewer.md

The Production Safety Guardian

• Specialty: Configuration changes that cause outages
• Focus: Connection pools, timeouts, memory limits
• Use When: After any code changes, especially config
• Unique Feature: Categorizes issues by production risk

Example output:

🚨 CRITICAL: Database pool size reduced from 20 to 5
This will cause connection starvation under normal load.
Recommendation: Test with production traffic first.

🏗️ backend-architect.md

The System Designer

• Specialty: API design and service boundaries
• Focus: RESTful APIs, microservices, database schemas
• Use When: Designing new services or major refactors
• Unique Feature: Contract-first API design

🔐 security-auditor.md

The Security Specialist

• Specialty: OWASP Top 10, auth flows, encryption
• Focus: JWT, OAuth2, input validation, SQL injection
• Use When: Security reviews, auth implementation
• Unique Feature: Severity-based vulnerability reports

🎯 prompt-engineer.md

The AI Optimizer

• Specialty: LLM prompt optimization
• Focus: Chain-of-thought, few-shot examples, model-specific tuning
• Use When: Building AI features, improving responses
• Unique Feature: Always shows complete, copy-pasteable prompts

📝 SEO Workflow (7 agents)

The Content Marketing System

Complete SEO workflow from strategy to optimization:

• seo-content-planner.md → Strategy and topic clusters
• seo-content-writer.md → SEO-optimized content creation
• seo-keyword-strategist.md → Keyword analysis and density
• seo-meta-optimizer.md → Titles, descriptions, schemas
• seo-structure-architect.md → Content organization
• seo-content-auditor.md → Quality and E-E-A-T analysis
• seo-authority-builder.md → Trust signals and credibility

🚀 How to Use This Repository

1. For Your Projects

Copy CLAUDE.md to your project root and customize:

cp CLAUDE.md /path/to/your/project/CLAUDE.md
# Edit to match your project's commands and architecture

2. For Specialized Tasks

Reference subagents in Claude Code conversations:

"Act as the code-reviewer subagent from my AI library and review this pull request"

3. For Prompt Engineering

Use prompts from the library as starting points:

# From prompt-engineer.md
Always display complete prompt text in clearly marked sections.
Never describe a prompt without showing it.

4. For Team Workflows

Share specific subagents with team members:

"Use the security-auditor approach for all auth-related PRs"
"Run backend-architect analysis before system design meetings"  

💡 Pro Tips

Combining Subagents

"Use code-reviewer for security issues, then backend-architect for API design recommendations"

Context Switching

"Switch to seo-content-writer mode and create a blog post about [topic]"

Project-Specific Adaptation

"Act as code-reviewer but focus on our GraphQL schema changes and Redis caching patterns"

📚 Featured Prompts

PM Copilot

• Socratic questioning for product decisions
• First-principles thinking frameworks
• Tradeoff analysis with logical reasoning
• Integration with analytics and project tools

System Prompts

• Writing style replication
• Video AI generation (Sora, Veo3)
• 3D rendering from reference images

🔄 Workflow Examples

Code Review Workflow

1. Write/modify code
2. Invoke code-reviewer subagent
3. Get severity-categorized feedback
4. Address critical issues first
5. Implement suggestions

Content Creation Workflow

1. seo-content-planner → Strategy
2. seo-keyword-strategist → Keywords
3. seo-content-writer → Content
4. seo-meta-optimizer → Metadata
5. seo-content-auditor → Quality check

System Design Workflow

1. backend-architect → Service boundaries
2. security-auditor → Security review
3. performance-engineer → Optimization
4. code-reviewer → Implementation review

🤝 Contributing

This is a living library. Contributions welcome:

• New Subagents: Follow the YAML frontmatter format
• Prompt Improvements: Test with real scenarios first
• Documentation: Share your workflow discoveries
• Examples: Add production use cases

🎓 Learning Resources

• CLAUDE.md Best Practices: See examples in this repo
• Subagent Design: Study existing agents for patterns
• Prompt Engineering: Check prompt-engineer.md for techniques
• AI Workflows: Explore prompt library/ for inspiration

---

Ready to 10x your AI productivity? Start with CLAUDE.md in your next project and experiment with the subagents. Your future self will thank you.

Built for the AI-native workflow. Optimized for Claude Code.