Vscode Copilot Instructions

VSCode Copilot Custom Instructions & Prompt Files

Expert guidance for creating and managing VSCode Copilot customization files including custom instructions, prompt files, and agent configurations.

When to Use This Skill

Activate this skill when:

• Creating .github/copilot-instructions.md files
• Building .instructions.md files with applyTo patterns
• Setting up AGENTS.md configurations
• Creating .prompt.md reusable prompts
• Configuring YAML frontmatter for any of these file types
• Understanding variable syntax (${variable})
• Referencing tools (#tool:toolName) or files in instructions
• Troubleshooting instruction file activation issues
• Sharing custom instructions across teams or repositories

Quick Overview: File Types

1. .github/copilot-instructions.md

Always-on workspace instructions — Applied automatically to all chat requests in the workspace.

When to use: General coding guidelines, team conventions, project-specific rules that apply throughout the project.

Location: .github/copilot-instructions.md at workspace root

Example:

# Project Guidelines

- Use TypeScript for all new code
- Follow functional programming patterns
- Write tests using Jest
- API responses must use our standard error format

📄 Complete Guide: references/copilot-instructions-guide.md

2. .instructions.md Files

Conditional, targeted instructions — Apply to specific file types, locations, or tasks using applyTo glob patterns.

When to use: Language-specific guidelines, framework-specific patterns, file-type-specific rules.

Location: .github/instructions/ folder or user profile

Example:

---
name: python-coding-standards
description: Python coding standards and best practices
applyTo: "**/*.py"
---

# Python Standards

- Follow PEP 8 style guide
- Use type hints for all functions
- Write docstrings using Google style

📄 Complete Guide: references/instructions-files-guide.md

3. AGENTS.md File

Multi-agent workspace instructions — Useful when working with multiple AI agents, applied automatically like copilot-instructions.md.

When to use: When you work with multiple AI coding assistants and want unified instructions.

Location: Workspace root, or subfolders (experimental)

Example:

# Agent Guidelines

All agents working in this codebase must:

- Never modify files in the `legacy/` directory
- Always run tests before committing
- Use the project's error handling utilities

📄 Complete Guide: references/agents-md-guide.md

4. .prompt.md Files

Reusable, on-demand prompts — Triggered via /command in chat for specific development tasks.

When to use: Standardized workflows, code generation templates, review checklists, common development tasks.

Location: .github/prompts/ folder or user profile

Example:

---
name: create-react-component
description: Generate a React component with TypeScript and tests
argument-hint: component-name
agent: edit
---

# Create React Component

Create a new React component named ${input:componentName} with:

1. TypeScript interface for props
2. Function component implementation
3. Jest test file with basic tests
4. Storybook story file

Follow our component patterns in [docs/component-guidelines.md](docs/component-guidelines.md).

📄 Complete Guide: references/prompt-files-guide.md

Decision Tree: Which File Type?

Do you want instructions that apply automatically?
├─ YES → Always-on for entire workspace?
│   ├─ YES → Use .github/copilot-instructions.md
│   └─ NO → Apply to specific files/patterns?
│       └─ Use .instructions.md with applyTo pattern
│
└─ NO → Want to trigger on-demand?
    └─ Use .prompt.md file (invoked via /command)

Also consider: Use AGENTS.md if working with multiple AI agents simultaneously.

Common Elements Across All File Types

YAML Frontmatter (Optional but Recommended)

All file types support YAML frontmatter for metadata:

---
name: file-name-slug
description: What this file does
applyTo: "**/*.py"           # .instructions.md only
agent: edit                  # .prompt.md only
tools: [githubRepo, files]   # .prompt.md only
---

Variable Syntax

Reference dynamic values in file bodies:

- `${workspaceFolder}` — Workspace root path
- `${file}` — Currently open file path
- `${selection}` — Selected text
- `${input:variableName}` — User input from chat
- `${input:variableName:placeholder}` — With placeholder hint

📄 Complete Reference: references/variables-and-references.md

Tool References

Reference agent tools using special syntax:

Use #tool:githubRepo to search the repository.
Use #tool:files to read workspace files.

File References

Link to other workspace files using relative Markdown links:

Follow the patterns in [docs/api-guide.md](docs/api-guide.md).
See [scripts/validate.py](scripts/validate.py) for examples.

Reference Documentation

All detailed guides are in the references/ folder:

• Copilot Instructions Guide — .github/copilot-instructions.md complete reference
• Instructions Files Guide — .instructions.md with applyTo patterns
• AGENTS.md Guide — Multi-agent workspace configuration
• Prompt Files Guide — .prompt.md reusable prompts
• Variables & References — All variable syntax and file referencing
• Best Practices — Writing effective instructions
• Troubleshooting — Common issues and solutions

Template Files

Ready-to-use templates in assets/templates/:

• copilot-instructions.template.md
• instructions-file.template.md
• agents.template.md
• prompt-file.template.md

Quick Start Workflows

Creating Always-On Instructions

# 1. Create .github directory if needed
mkdir -p .github

# 2. Create copilot-instructions.md
code .github/copilot-instructions.md

# 3. Add your guidelines
# 4. Enable in settings: github.copilot.chat.codeGeneration.useInstructionFiles

Creating Conditional Instructions

# 1. Create instructions directory
mkdir -p .github/instructions

# 2. Create instruction file
code .github/instructions/my-rules.instructions.md

# 3. Add YAML frontmatter with applyTo pattern
# 4. Enable: chat.includeApplyingInstructions

Creating Reusable Prompts

# 1. Create prompts directory
mkdir -p .github/prompts

# 2. Create prompt file
code .github/prompts/my-task.prompt.md

# 3. Use in chat: /my-task

Settings to Enable

Ensure these VS Code settings are enabled:

{
  // Enable copilot-instructions.md
  "github.copilot.chat.codeGeneration.useInstructionFiles": true,
  
  // Enable .instructions.md with applyTo patterns
  "chat.includeApplyingInstructions": true,
  
  // Enable AGENTS.md
  "chat.useAgentsMdFile": true,
  
  // Enable nested AGENTS.md in subfolders (experimental)
  "chat.useNestedAgentsMdFiles": false,
  
  // Show prompts as recommendations
  "chat.promptFilesRecommendations": true
}

Key Best Practices

✅ Keep Instructions Concise: Each instruction should be a single, clear statement
✅ Use Specific Examples: Show code examples of what you want
✅ Reference, Don't Duplicate: Use Markdown links to existing docs
✅ Test Activation: Verify instructions trigger correctly
✅ Layer Appropriately: Use the right file type for the right scope

❌ Avoid Vague Guidance: "Write good code" → Instead: "Use functional components with TypeScript interfaces"
❌ Don't Overload: Too many instructions reduce effectiveness
❌ Don't Forget Settings: Instructions won't work without proper VS Code settings

📄 Full Best Practices: references/best-practices.md

Troubleshooting

Instructions not applying?

1. Check the file is in the correct location
2. Verify the relevant setting is enabled
3. Check YAML frontmatter syntax
4. For .instructions.md: verify applyTo glob pattern matches
5. View diagnostics: Right-click in Chat → Diagnostics

📄 Full Troubleshooting Guide: references/troubleshooting.md

Example Use Cases

Use Case 1: Team Coding Standards

Solution: .github/copilot-instructions.md with always-on guidelines

Use Case 2: Python-Specific Rules

Solution: .instructions.md with applyTo: "**/*.py"

Use Case 3: Generate API Endpoint

Solution: .prompt.md with /create-endpoint command

Use Case 4: Multiple AI Assistants

Solution: AGENTS.md at workspace root

Community Resources

• Official VSCode Docs
• Prompt Files Docs
• Awesome Copilot Repository — Community examples

Remember: Instruction files are most effective when they're specific, actionable, and well-organized. Start simple and iterate based on what works for your team.