Pydantic AI Skills for Claude Code

Production-ready Claude Code skills for building type-safe AI agents with Pydantic AI. Comprehensive reference implementations covering dependency injection, tool calling, structured outputs, streaming, multi-agent orchestration, and LLM evaluation patterns.

Why Use These Skills?

• ✅ All examples tested against real LLMs - 34 integration tests validate every pattern with Claude Haiku 4.5 via OpenRouter. No broken snippets—guaranteed working code
• Battle-tested patterns - Real-world implementations, not toy examples
• Type-safe by design - Full Pydantic validation for inputs and outputs
• Multi-model support - Works with OpenAI, Anthropic, OpenRouter, and more
• Evaluation-driven - Built-in testing patterns for AI agent quality assurance
• Production-ready - Includes observability with Logfire integration

Skills Included

1. pydantic-ai-agents — Building AI Agents

Complete reference for building production AI agents with Pydantic AI:

Pattern	Description
Dependency Injection	Type-safe state management with dataclasses
System Prompts	Dynamic, context-aware prompt engineering
Tool Calling	Function tools with proper context handling
Structured Outputs	Pydantic model validation for LLM responses
OpenRouter Integration	Multi-model access (GPT-4, Claude, Llama, etc.)
Logfire Observability	Debugging, tracing, and monitoring
Response Streaming	Real-time token streaming
Multi-Agent Systems	Orchestrating specialized agent teams
Conversation Memory	Persistent history across turns

2. pydantic-evals — Testing AI Agents

Reference for evaluation-driven AI development:

Feature	Description
Type-Safe Datasets	Structured test case collections
Multiple Evaluators	Deterministic, LLM-as-Judge, custom, span-based
Logfire Integration	Trace-aware evaluation metrics
Best Practices	Evaluation-driven development (EDD) workflows

Installation

Using UV (Recommended)

# Clone the repository
git clone https://github.com/YOUR_USERNAME/pydantic-ai-skills.git
cd pydantic-ai-skills

# Install with UV
uv sync

Using pip

pip install -e ".[dev]"

Quick Start

Environment Setup

Create a .env file with your API keys (see .env.example):

OPENAI_API_KEY=your_key
OPENROUTER_API_KEY=your_key
LOGFIRE_API_KEY=your_key

Using as Claude Code Skills

Copy the skill directories to your Claude Code skills location:

# Copy skills to Claude Code
cp -r skills/pydantic-ai-agents ~/.claude/skills/
cp -r skills/pydantic-evals ~/.claude/skills/

Project Structure

├── skills/
│   ├── pydantic-ai-agents/
│   │   ├── SKILL.md              # Main skill documentation
│   │   └── references/           # 12 reference implementation files
│   └── pydantic-evals/
│       ├── SKILL.md              # Main skill documentation
│       └── references/           # Evaluator examples and guides
└── tests/                        # Comprehensive test suite

Development

Running Tests

All skill examples are covered by unit tests to ensure every code pattern works correctly. Tests run automatically on every PR via GitHub Actions.