In-Depth

Hands On with New Experimental GitHub Copilot 'Agent Skills' in VS Code

• By David Ramel
• 01/11/2026

Developers are always looking for ways to streamline workflows, automate repetitive tasks, and essentially make tools work smarter. With the latest update to Visual Studio Code, version 1.108 (December 2025), a powerful new feature has arrived that promises to do just that: experimental Agent Skills functionality in GitHub Copilot.

"VS Code now supports Agent Skills, allowing you to teach the coding agent new capabilities and provide domain-specific knowledge," the update post explains. "Agent Skills are folders of instructions, scripts, and resources that GitHub Copilot can load when relevant to perform specialized tasks." The functionality is based on an open standard that works across multiple AI agents, including GitHub Copilot in VS Code, GitHub Copilot CLI, and GitHub Copilot coding agent, the documentation states.

While it's a developer-centric feature, as a proof of concept I decided to put it to the test with a practical, non-coding use case. I write all my Visual Studio Magazine articles in VS Code because I like all the customization capabilities that let me tailor the environment to my specific needs, like spell checkers, formatting snippets triggered by keyboard shortcuts and so on. Sometimes I work on freelancer contributions or paste in text where I have to do several things to make our CMS happy and conform to our editorial style.

So for a PoC I created an Agent Skill to act as my personal copy editor, cleaning up common textual inconsistencies with a single command.

A Practical PoC: Cleaning Up Article HTML

To keep things simple at first, I just focused on three common issues:

• Curly Quotes and Apostrophes: Those aesthetically pleasing "curly" quotes and 'apostrophes' often cause issues, I'm told by our website czar, needing to be converted to "straight" ones.
• Em Dashes: The elegant long dash (--) frequently needs to be replaced with two hyphens (--) for compatibility.
• Double Spaces: Text from various sources often introduces rogue double spaces, which need to be collapsed into single spaces.

Instead of manually finding and replacing these elements in every single article, I envisioned a single command that would fix them all. That vision was realized by VS Code's new Agent Skills.

Setting Up Your First Agent Skill: A Streamlined Workflow

The process of creating an Agent Skill is surprisingly straightforward, designed to integrate seamlessly into your existing VS Code project structure. Here's how I set up my "straighten-quotes" skill, which is now evolving into a full-fledged formatting cleanup tool:

1. Enable Agent Skills in VS Code

First, ensure the feature is active. Open your VS Code Settings (the gear icon or Ctrl + ,), search for chat.useAgentSkills, and make sure the checkbox is selected. This tells VS Code to look for your custom skill definitions.

2. Create the Skill Folder Structure

Navigate to your project's root directory (in my case, my "Articles" OneDrive folder). Inside, create a specific folder structure that VS Code recognizes:

• Create a folder named .github.
• Inside .github, create another folder named skills.
• Inside skills, create a final folder for your specific skill. I named mine straighten-quotes because I started with just that to keep things super-simple and then added (and am adding) more transformations.

So, the final path looked like: YOUR_PROJECT_ROOT/.github/skills/straighten-quotes/.

3. Define the Skill with SKILL.md

The core of your Agent Skill is a markdown file named SKILL.md, placed directly inside your skill-specific folder (e.g., straighten-quotes). This file contains metadata about your skill and, crucially, the step-by-step instructions the AI agent will follow.

To create it:

• Select the straighten-quotes folder in the VS Code Explorer.
• Click the New File icon (the paper with a plus sign) at the top of the Explorer pane.
• Name the file exactly: SKILL.md.
• Once the empty SKILL.md file opens in the editor, paste in your instructions. Here's what my initial version looked like, and how it evolved:

---
name: straighten-quotes
description: Cleans up HTML text by fixing quotes, apostrophes, em dashes, and double spaces for CMS compatibility.
---

# Instructions

When I ask you to "clean this article" or "fix formatting," perform all these steps:

1. Replace all curly double quotes with straight double quotes.
2. Replace all curly single quotes and apostrophes with straight single quotes.
3. Find every long dash and replace it with two short dashes (--).
4. Remove any double spaces between words and replace them with a single space.

4. Putting the Skill to Work

With the SKILL.md file saved, to use it:

• Open any HTML article file in VS Code that needs formatting cleanup.
• Open the VS Code Chat panel (the speech bubble icon in the sidebar, or Ctrl + Shift + I).
• In the chat input, type a command that matches your skill's description. I typed: clean this article.

With some significant churning, including thinking, planning and internal discussions that I couldn't really keep track of, Copilot analyzed my request, identified the relevant skill, and applied the defined step. It then presented a clear diff of the proposed changes, allowing me to review and click "Keep" to implement the fixes or "Undo" if I wasn't happy with the diff.

Translating to Code: The Developer's Advantage

While my use case focused on text cleanup, the implications for developers are far-reaching. Agent Skills can automate virtually any oft-repeated task or incorporate project-specific knowledge directly into your AI assistant. Imagine tasks like:

• Code Refactoring: Automatically converting older syntax to newer standards across a codebase.
• API Call Standardization: Ensuring all new API calls follow a specific structure or include necessary authentication headers.
• Linting and Formatting: Enforcing project-specific style guides beyond what standard formatters do.
• Documentation Generation: Creating standard docstring templates for new functions or classes.
• Security Best Practices: Suggesting or enforcing common security patterns in new code.
• Debugging Assistance: Providing specific diagnostic steps for common errors in your project.
• Environment Setup: Automating initial setup steps for new team members.

The beauty of Agent Skills lies in their modularity and adaptability. You define the "how-to" once, and your AI assistant learns it permanently for that workspace. This transforms the AI from a general helper into a highly specialized, context-aware teammate.

Conclusion

The new Agent Skill functionality in VS Code is more than just a novelty; it's a significant leap towards truly personalized and intelligent coding assistance. By giving developers the power to "teach" their AI agent, VS Code has opened up a new realm of possibilities for automation, consistency, and ultimately, productivity. I encourage every developer to explore how Agent Skills can be tailored to their unique workflows -- you might be surprised at how much time you can save, and how much cleaner your code (or articles!) can become.