AI Tools & Claude Code Configuration

Personal repository for Claude Code CLI customization, AI experimentation, and enterprise development automation.

What's Inside

🤖 Custom Claude Code Agents

Specialized sub-agents for complex tasks:

meta-agent - Creates new sub-agents from descriptions
unit-test-writer - Generates comprehensive unit tests

🛠️ Skills & Commands

User-invocable skills (invoke with /skill-name):

backend-code-reviewer - C#/.NET code review with Microsoft standards
react-ux-code-reviewer - React/TypeScript UX and accessibility review
domain-driven-design-guru - DDD, Clean Architecture, and CQRS guidance
competency-tracker - Professional competency documentation
read-avro-files - Apache Avro file processing

Development workflows:

/plan-work - Create detailed implementation plans
/implement-plan - Execute plans with progress tracking
/create-adr - Generate Architecture Decision Records
/c4-documenter - Create C4 architecture diagrams
/new-worktree - Git worktree management

📚 AI Learning Journey

Structured 12-month learning plan covering:

Phase 1: GenAI & Enterprise Integration (LLMs, RAG, Agent frameworks)
Phase 2: ML/DL Fundamentals (Neural networks, transformers)
Phase 3: Advanced Topics (Multi-agent systems, AI safety)

See ClaudeCode/configuration/plans/ai-learning.md

⚙️ Configuration

settings.json - Permissions, hooks, and statusline config
statusline.ps1 - PowerShell statusline integration
notify-task-complete.ps1 - Task completion notifications

Quick Start

Using Custom Skills

claude /backend-code-reviewer    # Review C#/.NET code
claude /plan-work                # Plan implementation
claude /create-adr               # Create ADR

Creating New Agents

claude /meta-agent               # Launch agent creator

Agent Creation from MCP Tools

The AgentCreation/ folder contains documentation and samples demonstrating how to create custom agents that leverage MCP (Model Context Protocol) tools.

Workflow Overview

Discover available MCP tools - Ask Claude to list tools from your configured MCP servers (e.g., ElevenLabs, Firecrawl)
Identify target tools - Select the specific MCP tools you want your agent to use
Create agent via meta-agent - Provide a prompt describing what you want the agent to do and which tools it should use
Customize the output - Tweak the generated agent definition to fit your specific needs

Folder Contents

File	Description
`ReadMe.md`	Step-by-step guide with screenshots
`meta-agent-creator.md`	Sample meta-agent that creates new sub-agents
`work-summary-provider-agent.md`	Example agent using ElevenLabs MCP tools for TTS summaries

Example: Creating an Audio Summary Agent

The included work-summary-provider-agent.md demonstrates creating an agent that:

Uses mcp__ElevenLabs__text_to_speech to generate audio
Uses mcp__ElevenLabs__play_audio to play the result
Provides concise work completion summaries via audio

This pattern can be applied to any MCP server tools available in your configuration.

Key Tips

Use "Proactively" and "IMPORTANT" keywords in agent descriptions for automatic delegation
Be concise in agent prompts - aim for 2 sentences max when passing context
The meta-agent reads the latest Claude Code documentation to ensure proper format

Source: Claude Code Hooks Mastery

MCP Context Bloat Alternatives

MCP servers can bleed context tokens through auto-discovery before your agent starts working. The MCPContextBloatAlternative/ folder documents four approaches to agent tooling with different context/complexity trade-offs.

Approach	Context Cost	Best For
MCP Server	High	Multi-agent scale, standardized protocols
CLI as Tools	Medium	Build once, use everywhere
Scripts (Progressive Disclosure)	Low (~90% savings)	Context-critical workloads
Skills	Low	Claude Code ecosystem

The 80/10/10 Rule: Start with CLI (80%), extend to scripts when context matters (10%), wrap in MCP only for multi-agent scale (10%).

Source: IndyDevDan - Beyond MCP

AI Learning Projects

Organize projects under:

claude/tasks/ai-learning-progress/phaseN/
├── experiments/         # Proof-of-concept code
├── projects/           # Production-ready implementations
└── *.md               # Documentation and notes

Tech Stack Focus

Backend: C#, .NET, Azure, Domain-Driven Design
Frontend: React, TypeScript, Material-UI
AI/ML: Semantic Kernel, LangChain, Azure OpenAI, ML.NET
Tools: Claude Code CLI, PowerShell, Git

Documentation

CLAUDE.md - Comprehensive guide for Claude Code instances
ClaudeCode/configuration/ - All configuration files
AgentCreation/ - Guide for creating agents from MCP tools
MCPContextBloatAlternative/ - Alternatives to MCP for context-efficient tooling

Repository Structure

ai-tools/
├── AgentCreation/              # Agent creation guide & samples
├── MCPContextBloatAlternative/ # MCP alternatives for context efficiency
├── ClaudeCode/configuration/
│   ├── agents/                 # Custom sub-agents
│   ├── commands/               # Workflow commands
│   ├── skills/                 # User-invocable skills
│   ├── plans/                  # Planning documents
│   └── prompts/                # Prompt templates
├── CLAUDE.md                   # Claude Code guidance
└── README.md                   # This file

Note: This is a personal development repository focused on Claude Code customization and AI learning experimentation.

Name	fork-terminal-skill
Description	Fork a terminal session to a new terminal window. Use this when the user requests 'fork terminal' or 'create a new terminal' or 'new terminal: <command>' or 'fork session: <command>'.

fork-terminal-skill

SKILL.md