GitHub Copilot now supports SKILL.md through its agent mode, marking a major upgrade for developers who want to customize and extend Copilot’s capabilities with task-specific workflows. This integration enables developers to create and share modular skills—defined as markdown-based instructions—that Copilot can automatically recognize and execute depending on the context. In this guide, we’ll explore how to install these SKILL.md files, how Copilot chooses which skill to use, and ways teams can collaborate on skills to boost productivity across projects.

What Are SKILL.md Files?

SKILL.md files are structured markdown documents that define specific workflows or tasks for AI coding assistants. They provide clear, step-by-step instructions, examples, and parameters that help the AI understand exactly how to perform certain operations, such as generating code snippets, reviewing pull requests, or enforcing coding standards. These files are supported by multiple AI agents including Claude Code, Cursor, Codex CLI, and now GitHub Copilot through its agent mode.

How GitHub Copilot Uses SKILL.md

As of April 2026, GitHub Copilot’s agent mode can parse and execute tasks described in SKILL.md files. When you prompt Copilot to perform a specific task, it scans available SKILL.md files to find the most relevant skill to apply. This means Copilot is no longer limited to generic code suggestions based purely on its training data or the `.github/copilot-instructions.md` context file. Instead, it can follow precise, custom instructions tailored for your project or workflow.

Installing SKILL.md Skills in Your Repository

To start using SKILL.md skills with GitHub Copilot, you need to add SKILL.md files to your project repository. Follow these steps: 1. Create a folder named `.github/skills/` at the root of your GitHub repository. 2. Add your SKILL.md files into this folder. Each file should contain a clear description of the skill, usage instructions, and examples. 3. Commit and push these changes to your repository. Once pushed, Copilot agent mode will automatically detect these skills when you use Copilot in VS Code or on GitHub.com and suggest actions based on the instructions in the SKILL.md files.

Using Global Skills Across Projects

If you want to have skills available across multiple projects without copying SKILL.md files into each repository, you can place them in your global Copilot configuration directory. On your local machine, this is usually located at `~/.config/copilot/skills/` or a similar path depending on your OS. Adding SKILL.md files here makes these skills accessible to Copilot in any project you open, enabling consistent workflows and standards without the need for duplication.

How Copilot Chooses Which Skill to Use

Copilot analyzes your prompts and the context of your current code to determine which SKILL.md file is most relevant. It uses keyword matching, intent recognition, and the structure of your request to pick the best skill. For example, if you ask Copilot to "refactor this function according to our style guide," and you have a refactoring skill defined in `.github/skills/refactor.SKILL.md`, Copilot will apply that skill. If multiple skills seem relevant, Copilot prioritizes skills defined in the current repository over global skills, ensuring project-specific instructions take precedence. If no matching skill is found, Copilot falls back to its default behavior based on training and copilot-instructions.md context.

Practical Example: Creating a Code Review Skill

Imagine your team wants to automate code review comments for common issues. You can create a SKILL.md file named `code-review.SKILL.md` inside `.github/skills/` with instructions like: - Look for functions longer than 50 lines and suggest splitting. - Check for missing docstrings and recommend adding them. - Verify naming conventions according to your style guide. When reviewing a pull request, you prompt Copilot with "Review this code for style and best practices," and it uses the skill to generate detailed feedback, saving reviewers time and improving code quality.

Step-by-Step: Adding and Using a Skill

1. Write your SKILL.md file with clear instructions and examples. 2. Place the file in `.github/skills/` or your global skills folder. 3. Commit and push the changes if using repository skills. 4. Open your project in VS Code or GitHub.com with Copilot agent mode enabled. 5. Ask Copilot to perform a task related to your skill, such as "Generate unit tests for this module." 6. Copilot reads the SKILL.md, understands the task, and provides tailored suggestions. This workflow helps teams standardize complex or repetitive tasks by encoding them in skill files.

Tips for Writing Effective SKILL.md Files

- Use clear, concise language and stepwise instructions. - Include examples of inputs and expected outputs. - Document any parameters or options the skill can accept. - Keep skills focused on a single task for better maintainability. - Test your skill locally before sharing it across projects. - Use consistent naming conventions for skill files to make them easy to find. Well-written SKILL.md files improve Copilot’s accuracy and usefulness.

Team Collaboration and Sharing Skills

Teams can maintain a centralized skills repository that contains all approved SKILL.md files. This repository can be linked as a git submodule or referenced in CI/CD pipelines to ensure all projects use the latest skills. Developers can propose changes via pull requests, enabling peer review and continuous improvement of skills. Moreover, sharing skills reduces onboarding time for new developers by providing ready-made workflows and coding standards directly integrated into their coding environment.

Real-World Use Cases

- Enterprise Development: Large teams create skills for security checks, compliance validations, and code documentation enforcement. - Open Source Projects: Maintainers provide skills for contributors to generate tests, format code, or run benchmarks. - Individual Developers: Coders automate personal workflows like generating boilerplate code, setting up project scaffolds, or automating deployment scripts. For instance, a fintech company might have a skill that instructs Copilot to redact sensitive information before logging, ensuring compliance with data protection policies.

Conclusion

GitHub Copilot’s support for SKILL.md in agent mode unlocks new possibilities for customizing AI-assisted development. By defining clear, reusable skills, developers and teams can streamline workflows, enforce standards, and share best practices seamlessly across projects. Whether you’re automating code reviews, generating tests, or managing complex workflows, SKILL.md files make your AI assistant smarter and more aligned with your unique development needs. Start experimenting with skills today by adding them to your `.github/skills/` folder or your global config, and watch Copilot transform into a powerful, context-aware coding partner.