AI Agents  

Part 3 — Using skills.md Files with C# AI Agents

AI Code Review Image

Introduction

In Part 2, we built a modular AI Agent architecture with reusable skills.

However, there was still one major problem:

Prompt logic was hardcoded in C#.

This becomes difficult to maintain as the number of skills grows.

Modern AI Agent systems solve this by moving prompts into external markdown-based skill files.

This creates:

  • Cleaner architecture

  • Reusable skills

  • Versioned prompts

  • Easier prompt engineering

  • Non-developer customization

In this article, we’ll connect skills.md files to our C# AI Agent.

What Is skills.md?

A skills.md file defines:

  • Purpose

  • Rules

  • Instructions

  • Examples

  • Expected outputs

Example:

# Code Review Skill

You are a senior C# code reviewer.

Review ONLY the changed code.

Focus on:
- Bugs
- Security
- Performance

Git Diff:

{{GIT_DIFF}}

This becomes a reusable AI capability.

Recommended Folder Structure

/Skills
   /code-review.md
   /security-review.md
   /fix-generator.md

Step 1 — Load Skill File

public class SkillLoader
{
    public static string LoadSkill(string skillName)
    {
        string path = Path.Combine(
            AppDomain.CurrentDomain.BaseDirectory,
            "Skills",
            $"{skillName}.md");

        return File.ReadAllText(path);
    }
}

Step 2 — Replace Variables

public static class PromptBuilder
{
    public static string Build(
        string template,
        Dictionary<string, string> variables)
    {
        foreach (var variable in variables)
        {
            template = template.Replace(
                $"{{{{{variable.Key}}}}}",
                variable.Value);
        }

        return template;
    }
}

Step 3 — Build Dynamic Prompt

string template =
    SkillLoader.LoadSkill("code-review");

string diff =
    await GitHelper.GetDiffAsync();

string prompt =
    PromptBuilder.Build(
        template,
        new Dictionary<string, string>
        {
            ["GIT_DIFF"] = diff
        });

Now prompts are completely externalized.

Step 4 — Send to LLM

string result =
    await aiService.ReviewAsync(prompt);

Console.WriteLine(result);

Benefits of skills.md

Reusable Skills

The same skill can be used by:

  • VS Code

  • PR Review

  • CLI Tools

  • CI/CD Pipelines

Easier Prompt Engineering

Prompt updates no longer require code changes.

Better Agent Planning

Agents can dynamically select skills.

Example:

If auth files changed
    → Use security-review.md

Versioning

Prompts become version-controlled assets.

Recommended Enterprise Structure

/Skills
   /CodeReview
      skill.md
      examples.md
      config.json

   /SecurityReview
      skill.md

   /FixGenerator
      skill.md

Final Thoughts

Across this 3-part series, we built a complete evolution path:

Part 1

  • Git Diff → LLM → Review

Part 2

  • AI Agent → Skills → Review + Fix

Part 3

  • skills.md → Dynamic AI Behavior

This architecture provides:

  • Modularity

  • Scalability

  • Maintainability

  • Reusable AI capabilities

The future of developer tooling is moving toward intelligent AI Agents that:

  • Understand repositories

  • Review code

  • Generate fixes

  • Validate outputs

  • Collaborate with developers

And C# provides an excellent ecosystem for building these systems at enterprise scale.

Summary

This article demonstrated how to externalize AI prompts using skills.md files in a C# AI Agent architecture. By moving prompt logic outside the application code, developers can build modular, reusable, maintainable, and scalable AI systems with better prompt management, easier updates, and improved enterprise-level flexibility.

Membership not found