Copilot Agents
Why Copilot Agents
Customized instructions, prompts, agent skills, agent MCPs and custom agents help guide GitHub Copilot by providing contextual details about your repository such as the type of workflow your team follows, tools and other project specific details such as coding style, frameworks used or project specific rules.
Tip: Learn more about customizing GitHub Copilot in VS Code in the VS Code documentation.
Instructions
Instructions provides Copilot with repository-specific context, such as coding standards, frameworks, or workflows, to improve code suggestions.
Boilerplates & Templates
Standard language template to build instruction files.
Standard tool boilerplate for infrastructure-as-code tools.
Templates
Boilerplate
Language & Stack
System libraries, CLI tools, and embedded applications with POSIX/GNU libc.
.NET applications with modern C# patterns and best practices.
Modern C++ development with STL, RAII, and performance optimization.
Go for microservices, CLI tools, and concurrent applications.
Enterprise Java development with Spring framework and modern patterns.
Modern JavaScript with ES6+, Node.js, and browser development.
Kotlin for Android development and multi-platform projects.
Lua scripting for embedded systems, game development, and automation.
Python development for web applications, data science, and automation.
Systems programming with Rust's ownership model and memory safety.
iOS and macOS development with Swift and SwiftUI.
TypeScript development for web and Node.js applications.
C
C-Sharp
C++
Go
Java
JavaScript
Kotlin
Lua
Python
Rust
Swift
TypeScript
Framework / Library
Interactive terminal applications with Charm's Bubble Tea framework and Golang Cobra CLI.
Azure Function Apps using TypeScript Node.js.
REST API development with Express.js and TypeScript Node.js.
Cobra CLI (Go)
Node.js (TypeScript)
Tools
Drupal 11 module and theme development.
Standard Terraform instructions.
Terraform workflow orchestration with Atmos framework.
Content Management Systems (CMS)
Drupal
Infra as Code (IaC)
Terraform
Workflows
Create detailed Product Requirements Documents.
Break PRDs into actionable development tasks.
Systematic task execution with proper testing and Git practices.
AI Development Instructions
A comprehensive workflow for AI-assisted development featuring structured approaches to planning, task generation, and execution.
Prompts
Prompts are reusable tasks or workflow instructions that help guide Copilot to perform specific actions or generate certain outputs.
AI Development Tasks
Create detailed Product Requirements Documents using prompt tasks.
Break PRDs into actionable development tasks using prompt tasks.
Systematic task execution with proper testing and Git practices using prompt tasks.
A comprehensive workflow for AI-assisted development featuring structured approaches to planning, task generation, and execution.
Custom Agents
Custom agents let you set up different AI personas in VS Code for specific dev roles (like security reviewer, planner, or architect), each with its own instructions, tools, and behavior. You can also use handoffs to move between these specialized agents in a guided workflow (e.g., planning → implementation → review) with relevant context carried over.
The built-in available custom agents are:
- Agent
- Ask
- Edit
- Plan
- AIAgentExpert
- Configure Custom Agents (create your own)
AI Development Mode
Design and plan software systems.
Write clean, readable, and maintainable code using clean code best practices.
Debug your application code to find a fix.
Build Product Requirements Document (PRD).
Agent Skills
Agent Skills are portable, open standard, version-controlled folders of instructions, scripts, and resources that agents can discover and load on demand to do tasks more accurately and efficiently. They let agents gain domain expertise, new capabilities, and repeatable workflows—while making those same skills reusable across different compatible agent products and teams.
General
Performs arbitrary-precision arithmetic calculations including addition, subtraction, multiplication, division, and exponents.
Interact with Jira from the command line to create, list, view, edit, and transition issues, manage sprints and epics, and perform common Jira workflows.
Create new skills, modify and improve existing skills, and measure skill performance.
Documents
Document creation, editing, and analysis with support for tracked changes, comments, formatting preservation, and text extraction.
PDF manipulation toolkit for extracting text and tables, creating new PDFs, merging/splitting documents, and handling forms.
Presentation creation, editing, and analysis.
Spreadsheet creation, editing, and analysis with support for formulas, formatting, data analysis, and visualization.
Cloud
Azure CLI documentation to execute or ask about Azure CLI commands.
Look up and compare Azure service pricing using the Azure Retail Prices API.
Development
Automate browser interactions, test web pages and work with Playwright tests.
Create distinctive, production-grade frontend interfaces with high design quality.
Toolkit for interacting with and testing local web applications using Playwright.
MCPs
MCPs (Model Context Protocol servers) give agents a standardized way to connect to external tools, APIs, data sources, and local capabilities. They extend what an agent can do beyond plain chat by exposing actions such as reading files, browsing the web, querying cloud platforms, or interacting with development tooling through a common protocol.
This section highlights useful MCP servers you can add to your Copilot setup to expand agent capabilities for research, development, automation, and cloud workflows.
General MCPs
Create a local knowledge graph for your agent to remember information across different chat sessions.
Read and write files in batch operations, search file contents, list files, and validate file paths.
Automate web browsing for agents to retrieve and process content from web pages, converting HTML to markdown for easier consumption.
Break down complex problems into structured steps.
Allow your agent access to repository and workflow management.
Enables agents to get current time information and perform timezone conversions using IANA timezone names, with automatic system timezone detection.
Development MCPs
Playwright MCP to automate browser interactions, test web pages and work with Playwright tests.
Inject version-specific code documentation in your agent session to provide the correct API docs for code generation.
Cloud MCPs
Azure MCP Server supercharges your agents with Azure context across different Azure services.
Agent tools to access AWS documentation, search for content, and get recommendations.
Agent tools to interact with the Google Cloud environment using the gcloud CLI.
MCP server bridging AI agents to multi-cluster Kubernetes environments for cluster management, pod inspection, and real-time observability.
How to Use
Setup Copilot in VSCode
- Hover over the Copilot icon in the Status Bar and select Set up Copilot.
- Select Sign in to sign in to your GitHub account or Use Copilot if you're already signed in.
Tip: Read more about setting up VS Code Copilot.
Setup Instructions
- Create instruction files using the latest naming conventions:
- Workspace instructions (place
*.instructions.mdfiles in.github/instructions/directory). - Workspace prompts (place
*.prompt.mdfiles in.github/prompts/directory). - Workspace custom agents (place
*.agent.mdfiles in.github/agentsdirectory). - Workspace custom skills (skills are stored in directories with a
SKILL.mdfile that defines the skill's behavior). - Workspace single instruction (place
copilot-instructions.mdin.githubdirectory).
- Workspace instructions (place
File Types
Instruction Files
.instructions.md - Contextual instructions that apply to specific files or file types.
Prompt Files
.prompt.md - Reusable prompts for specific tasks or workflows.
Custom Agent Files
.agent.md - Predefined AI personas behavior in VS Code for specific dev roles.
Custom Agent Skills
SKILLS.md - Portable, version-controlled folders of instructions, scripts, and resources that agents can discover and load on demand.
Formatting
Use YAML front matter to specify metadata like applyTo, mode, and description.