What Are GitHub Copilot Skills?

In the evolving landscape of AI-assisted development, GitHub Copilot Skills (also referred to as Agent Skills) represent a significant leap forward in customization. According to recent documentation from GitHub, these skills allow developers to modify Copilot's behavior and abilities when it works on particular tasks.

Functionally, GitHub Copilot Skills are folders containing instructions, scripts, and resources. When relevant, Copilot can load these assets to improve its performance in specialized tasks. This transforms Copilot from a general-purpose coding assistant into a specialized agent capable of executing complex, multi-step procedures defined by you.

Currently, these agent skills work with the Copilot coding agent, the GitHub Copilot CLI, and the agent mode in Visual Studio Code Insiders. They are designed for reusable workflows, making them ideal for DevOps, SREs, and teams that require standardized coding practices.

The Anatomy of an Agent Skill

Understanding the directory structure is crucial for implementing GitHub Copilot Skills effectively. A skill is not just a single file; it is a directory structure that houses the logic your agent needs.

.github/skills/
OR
.claude/skills/

~/.copilot/skills/
OR
~/.claude/skills/

SKILL.md

Detailed Component Breakdown

To create a functional GitHub Copilot Skill, you must adhere to specific naming conventions. Skill subdirectory names should be lowercase and use hyphens for spaces (for example, webapp-testing).

The SKILL.md file acts as the bridge between your intent and the AI's execution. When Copilot chooses to use a skill, this file is injected into the agent's context. This injection gives the agent access to your custom instructions, allowing it to follow specific guidelines, utilize bundled scripts, or reference example files you have included in the skill's directory.

How to Create GitHub Copilot Skills

Creating your first agent skill is a straightforward process. By following this step-by-step guide, you can start building a library of GitHub Copilot Skills to automate your daily tasks.

Step 1: Create the Skills Directory

First, determine if you want a project-specific skill or a personal skill. For a project-specific skill, navigate to your repository root and create a .github/skills directory. If you prefer a personal skill available everywhere, create a ~/.copilot/skills directory in your user home folder.

Step 2: Create a Subdirectory for Your Skill

Inside your skills folder, create a new directory for the specific task. Remember, the naming convention requires lowercase letters and hyphens.

mkdir .github/skills/image-conversion-workflow

Step 3: Create the SKILL.md File

Inside your new subdirectory (for example, image-conversion-workflow), create a file named exactly SKILL.md. This file must contain YAML frontmatter and a Markdown body.

---
description: A skill to convert SVG images to PNG format using local scripts.
license: MIT
---

# Image Conversion Instructions

When the user asks to convert images, use the script located at ./convert.sh.
Ensure that the output directory exists before running the conversion.
...

Step 4: Add Supplementary Resources

One of the most powerful features of GitHub Copilot Skills is the ability to bundle assets. You can add scripts, templates, or example files into the same subdirectory.

For example, if you were writing a skill for converting images between different formats, you might include a script for converting SVG images to PNG. Your SKILL.md instructions should then explicitly tell Copilot when and how to use these resources.

Using Skills in Chat & CLI

Once you have defined your GitHub Copilot Skills, there are two primary ways to interact with them: Automatic Discovery and Manual Invocation.

Automatic Discovery

When performing tasks, Copilot is intelligent enough to decide when to use your skills. This decision is based on your prompt and the skill's description defined in the YAML frontmatter. If your request matches the intent of a skill, Copilot will automatically load the SKILL.md context.

Manual Invocation (CLI & Chat)

You can also explicitly tell Copilot to use a specific skill. In the GitHub Copilot CLI or Chat interface, this is often done via slash commands.

• List available skills: You can type /skills list or simply ask, "What skills do you have?"
• Toggle skills: Use the command /skills to see an interactive list where you can use the up and down keys to toggle specific skills on or off.
• Direct Invocation: As noted in community guides, you can invoke a skill manually by typing / in the chat input, picking the skill, and adding extra context (for example, /incident-triage payments-api prod).

Platform Support

It is important to note the current compatibility. GitHub Copilot Skills currently work with:

• GitHub Copilot CLI
• Copilot Coding Agent
• VS Code Insiders (Agent Mode)

Note: Support in the stable version of VS Code is announced to be coming soon.

Why Invest in Agent Skills?

Adopting GitHub Copilot Skills transforms the way teams interact with AI. Here are the core benefits and use cases:

Frequently Asked Questions

Common questions regarding GitHub Copilot Skills and agent configuration.

Where should I store my skills?

For repository-specific skills, use .github/skills or .claude/skills. For personal skills shared across your environment, use ~/.copilot/skills or ~/.claude/skills.

What file format is required?

Every skill folder must contain a SKILL.md file. This file combines Markdown for instructions and YAML frontmatter for metadata like descriptions and licenses.

Can Copilot use external scripts?

Yes. Agent skills are folders. You can include scripts (for example, Python or Bash) inside the skill directory. Your SKILL.md should instruct Copilot on when and how to execute these scripts.

How does Copilot know which skill to use?

Copilot uses semantic matching based on your prompt and the description provided in the skill's SKILL.md file. You can also manually invoke them using slash commands in supported interfaces.

References

• GitHub Copilot Docs
• OpenClaw Skills Docs
• OpenClaw CLI Skills Docs
• How to Install OpenClaw Skills
• OpenClaw Skill Troubleshooting: 15 Common Errors and Fixes