AI-Toolkit: AI Programming Skills Library for Claude Code and OpenCode

AI-Toolkit is a comprehensive collection of AI programming skills supporting two major platforms—Claude Code and OpenCode. It covers professional skills in multiple domains including TDD, .NET development, edge computing, RAG, MCP, and software engineering judgment training. Its core design is a two-layer architecture of Skills and Subagents, aiming to improve the efficiency and quality of AI-assisted software development, emphasizing the positioning of AI as a tool to enhance developers' capabilities rather than a replacement.

With the rise of AI programming tools like Claude Code, OpenCode, and Cursor, developers face new challenges in how to effectively utilize these tools. AI-Toolkit emerged as a systematic collection of skills, achieving cross-platform compatibility through standardized AGENTS.md files, avoiding reinventing the wheel, and solving the problem of skill reuse between different tools.

AI-Toolkit adopts a layered architecture, divided into two levels: Skills and Subagents:

• Skills: Domain-specific knowledge bases and execution protocols (e.g., TDD-cycle defines the red-green-refactor process);
• Subagents: AI agents that autonomously execute tasks and call skills to complete complex workflows (e.g., code-review-agent uses the automated-code-review skill). This design achieves separation of concerns—skills focus on 'what to do/how to do it', while subagents focus on 'when to do it/coordinate execution'—balancing flexibility and maintainability.

AI-Toolkit covers multiple domain skill sets:

1. TDD Skill Set: Based on Kent Beck's principles, includes 6 core skills such as tdd-cycle and tdd-implementer, covering the complete TDD lifecycle;
2. .NET Skill Set: Includes enterprise-level development skills like vertical slice architecture, EF migration management, and security reviews (OWASP and federal compliance);
3. Edge Computing Skill Set: Optimized for Jetson/RPi, including computer vision pipelines and sensor integration;
4. RAG and Local LLM Skill Set: Provides Python/.NET RAG scaffolding, MCP server setup, Ollama model management, etc.

The uniqueness of AI-Toolkit lies in its judgment training skill set, which cultivates developers' core competitiveness through deliberate practice:

• architecture-review: Challenges designs with Socratic questioning (from angles like SOLID and coupling);
• pattern-tradeoff-analyzer: Analyzes pattern tradeoffs to avoid the 'golden hammer' tendency;
• system-design-kata: Domain calibration exercises (e.g., security workflows, edge clusters);
• security-review-trainer: Progressive vulnerability discovery training. These skills go beyond code generation and strengthen human engineering judgment.

AI-Toolkit represents a new paradigm for AI-assisted development: it does not replace developers but enhances their capabilities through structured skills. It acknowledges the value of AI tools while emphasizing the importance of human judgment, engineering practices, and domain knowledge. For Claude Code/OpenCode users, it provides a reusable skill library to improve efficiency and demonstrates the 'AI augments humans' thinking framework.