The AI Agent Skills

Introducing CrestApps.AgentSkills: A Structured Foundation for AI Skills in .NET

AI integration in .NET applications is evolving rapidly. But without structure, it can easily become a collection of scattered prompts, hardcoded tool calls, and tightly coupled logic.

The CrestApps.AgentSkills ecosystem provides a clean, layered architecture for defining, organizing, and exposing AI skills in .NET — with first-class support for the Model Context Protocol (MCP) and OrchardCore.

This post walks through the projects in the repository and explains how they fit together — based strictly on their actual structure and purpose.

1. CrestApps.AgentSkills — The Foundation

At the core of the ecosystem is:

CrestApps.AgentSkills

This is the foundational .NET project that defines reusable AI skills and the infrastructure to work with them.

It provides:

A structured way to define AI agent skills
Common abstractions for skill execution
A shared engine that other packages build upon

This project is not OrchardCore-specific. It is designed for general .NET applications and acts as the base layer for everything else in the ecosystem.

If you are building AI-capable .NET applications and want a standardized way to define and organize skills, this is the starting point.

📦 Installation

Using the .NET CLI:

dotnet add package CrestApps.AgentSkills

Or using the Package Manager Console:

Install-Package CrestApps.AgentSkills

2. CrestApps.AgentSkills.Mcp — MCP Integration for .NET

Once skills are defined, they need to be discoverable and callable by external AI systems.

That is the role of:

CrestApps.AgentSkills.Mcp

This project integrates the core skill engine with the Model Context Protocol (MCP).

It enables:

Exposure of registered skills through MCP
Skill discovery by MCP-compatible clients
Invocation of .NET skills via protocol calls

In practical terms, this allows your .NET application to act as an MCP server. External AI agents and tools that support MCP can inspect available skills and call them in a structured way.

This shifts AI integration from prompt-based guessing to protocol-driven interaction.

📦 Installation

Using the .NET CLI:

dotnet add package CrestApps.AgentSkills.Mcp

Or using the Package Manager Console:

Install-Package CrestApps.AgentSkills.Mcp

Note: This package depends on CrestApps.AgentSkills and will install it automatically if it is not already referenced.

3. CrestApps.AgentSkills.OrchardCore — OrchardCore-Specific Skills

For developers working with OrchardCore, the repository includes:

CrestApps.AgentSkills.OrchardCore

This project extends the core skill engine with OrchardCore-specific functionality.

It provides:

OrchardCore-aware skill definitions
Extensions tailored to CMS concepts
Conventions for working with OrchardCore entities and structures

While the base package works for any .NET application, this project brings domain knowledge specific to OrchardCore into the skill system.

It enables AI skills that understand OrchardCore constructs rather than treating the CMS as a generic application.

📦 Installation

Using the .NET CLI:

dotnet add package CrestApps.AgentSkills.OrchardCore

Or using the Package Manager Console:

Install-Package CrestApps.AgentSkills.OrchardCore

Note: This package requires an OrchardCore application and depends on CrestApps.AgentSkills.

4. CrestApps.AgentSkills.Mcp.OrchardCore — OrchardCore MCP Runtime

Finally, the ecosystem includes:

CrestApps.AgentSkills.Mcp.OrchardCore

This package combines:

OrchardCore-specific skills
MCP server capabilities

It allows a running OrchardCore application to expose its skills through MCP at runtime.

With this in place, an OrchardCore instance can:

Publish its available skills
Respond to MCP calls from external agents
Participate in a broader AI tooling ecosystem

This creates a bridge between a live CMS environment and AI agents that can interact with it programmatically.

📦 Installation

Using the .NET CLI:

dotnet add package CrestApps.AgentSkills.Mcp.OrchardCore

Or using the Package Manager Console:

Install-Package CrestApps.AgentSkills.Mcp.OrchardCore

Note: This package depends on:

CrestApps.AgentSkills

CrestApps.AgentSkills.Mcp

CrestApps.AgentSkills.OrchardCore

All dependencies will be installed automatically by NuGet.

How the Projects Fit Together

The ecosystem is layered by design:

CrestApps.AgentSkills
Core engine for defining and executing AI skills in .NET.
CrestApps.AgentSkills.Mcp
Adds MCP protocol support to expose skills externally.
CrestApps.AgentSkills.OrchardCore
Adds OrchardCore-specific skill definitions.
CrestApps.AgentSkills.Mcp.OrchardCore
Hosts OrchardCore skills through MCP at runtime.

Each project builds directly on the previous layer, creating a clean separation of concerns.

Why This Architecture Matters

This ecosystem enables:

Structured AI skill definitions instead of ad-hoc prompts
Reusable skills across multiple projects
Protocol-based integration using MCP
Domain-aware AI capabilities for OrchardCore
Runtime exposure of application capabilities to AI agents

For .NET developers, this means moving from experimental AI integrations to maintainable, discoverable, and extensible AI systems.

Final Thoughts

The CrestApps.AgentSkills ecosystem provides a practical foundation for building AI-capable .NET and OrchardCore applications using structured skills and the Model Context Protocol.

Instead of embedding AI logic directly into application code in an ad-hoc way, you get:

A reusable skill engine
A protocol bridge for external agents
Domain-specific extensions
A runtime hosting model

If you're building AI-driven .NET systems — especially with OrchardCore — this architecture offers a clean and scalable approach.