A .NET implementation of the Agent Skills open standard by Anthropic.

AgentSkills.NET enables .NET-based AI agents to discover, load, and execute skills defined in a standardized format. It provides:

• Skill Discovery: Scan directories for skill definitions
• YAML + Markdown Parsing: Parse SKILL.md files with frontmatter and instructions
• Validation: Comprehensive validation with diagnostics (no exceptions for normal errors)
• Progressive Disclosure: List skills first, load full content only when activated
• Host-Agnostic Design: Core libraries work with any agent framework
• Safety-First: Scripts are treated as data; no execution by default

New to AgentSkills.NET? Start here: 📘 Getting Started Guide - Get up and running in 15 minutes!

# Clone and build (NuGet packages coming soon)
git clone https://github.com/sfmskywalker/agentskillsdotnet.git
cd agentskillsdotnet
dotnet build

using AgentSkills.Loader;
using AgentSkills.Validation;
using AgentSkills.Prompts;

// Load skill metadata (fast path - no full content)
var loader = new FileSystemSkillLoader();
var (metadata, diagnostics) = loader.LoadMetadata("./skills");

// Validate and filter
var validator = new SkillValidator();
var validMetadata = metadata
    .Where(m => validator.ValidateMetadata(m).IsValid)
    .ToList();

// Render list for LLM (progressive disclosure - stage 1)
var renderer = new DefaultSkillPromptRenderer();
var listPrompt = renderer.RenderSkillList(validMetadata);

// When LLM activates a skill, load full details (stage 2)
var (skill, _) = loader.LoadSkill("./skills/chosen-skill");
var detailsPrompt = renderer.RenderSkillDetails(skill);

# Run the walking skeleton sample (basics)
dotnet run --project samples/AgentSkills.Sample/AgentSkills.Sample.csproj

# Run the Microsoft Agent Framework integration demo (explains concepts)
dotnet run --project samples/AgentSkills.Sample.AgentFramework/AgentSkills.Sample.AgentFramework.csproj

# Run the full Microsoft Agent Framework sample (working tools + function calling)
dotnet run --project samples/AgentSkills.MicrosoftAgentFramework.Sample/AgentSkills.MicrosoftAgentFramework.Sample.csproj

🚧 Early Development - This project is in active development. APIs may change.

• 📘 Getting Started Guide - 15-minute quick start tutorial
• ✍️ Skill Authoring Guide - How to write excellent skills
• ❓ FAQ - Frequently asked questions

• 🔒 Security & Safety Guide - Security model and host guarantees
• 📋 Project Brief - Architecture and design principles
• 📝 Prompt Rendering Guide - Progressive disclosure patterns
• 💡 Usage Scenarios - Real-world integration examples

• 📚 Public API Reference - Complete API documentation
• 🧪 Testing Guide - Testing strategy and best practices
• 🔧 Troubleshooting Guide - Common issues and solutions
• 🔍 Reference Comparison - Comparison with Python reference implementation

• 🤝 Contributing Guide - How to contribute
• 🤖 Agent Guidelines - Guidelines for AI coding agents
• 📐 Architecture Decisions - ADRs documenting key decisions

├── src/                   # Core library source code
│   ├── AgentSkills/      # Core domain models
│   ├── AgentSkills.Loader/      # Skill loading
│   ├── AgentSkills.Validation/  # Validation rules
│   └── AgentSkills.Prompts/     # Prompt rendering
├── tests/                # Test projects
├── samples/              # Sample applications
├── docs/                 # Documentation
├── fixtures/             # Test fixtures and example skills
│   └── skills/          # Example skills demonstrating patterns
└── .github/workflows/    # CI/CD automation

The fixtures/skills/ directory contains example skills demonstrating various patterns:

Complete Examples:

• example-skill - Basic skill structure
• complete-skill - All optional fields and resources
• minimal-skill - Minimal valid skill
• email-sender - Real-world email composition skill
• code-reviewer - Systematic code review workflow

Edge Cases:

• special-chars-skill - Special characters and Unicode
• large-instructions-skill - Performance testing

See the fixtures README for a complete catalog.

# Restore dependencies
dotnet restore

# Build all projects
dotnet build

# Run tests
dotnet test

Contributions are welcome! Please read our Contributing Guide before submitting PRs.

Key points:

• Follow the conventions in AGENTS.md
• Keep changes small and focused
• Add tests for new functionality
• Update documentation as needed

1. Safety First: No script execution by default
2. Progressive Disclosure: Load metadata fast, full content on demand
3. Host Agnostic: Core libraries don't depend on specific frameworks
4. Diagnostics over Exceptions: Return diagnostics for validation errors
5. Explicit APIs: Boring, clear, discoverable interfaces

See docs/project_brief.md for details.

This project is licensed under the MIT License - see the LICENSE file for details.

• Agent Skills Standard by Anthropic
• The .NET and AI communities