GitHub Copilot Workflow Cheatsheet

Why This Cheatsheet?

GitHub Copilot has evolved far beyond simple code completions. With chat, inline editing, custom instructions, agents, and MCP servers, it's now a full-fledged AI coding companion. But with so many features, it's easy to miss the ones that could 10x your productivity.

This cheatsheet breaks down the essential Copilot workflows — from basic completions to advanced customization — so you can code faster, smarter, and with better context.

1. Copilot Architecture & Flow

Understanding how your prompt flows through Copilot helps you debug unexpected output and explain the system to stakeholders.\n

How Your Prompt Flows Through Copilot

💬 User Prompt (Chat · Inline · CLI · Completions)
         ↓
📂 Context Gathering (Open tabs · #file · #codebase · Instructions · Skills)
         ↓
🛡️ Pre-Model Filters (Responsible AI · Content Exclusion)
         ↓
🔀 Proxy Service (Routes to model · Applies policies)
         ↓
🧠 AI Model (GPT · Sonnet · Gemini · Opus · Grok — User-selected or auto)
         ↓
🔍 Post-Model Filters (Duplicate code check · Safety & quality)
         ↓
✅ Response delivered to user

Context Sources

Source	Description
Open Tabs	Active editor files
Workspace	`#codebase` search
Instructions	`.md` config files
Skills	`SKILL.md` bundles
MCP Servers	External tools/data
Selection	`#selection` context
Web Pages	`#fetch` URL content

Output Types

⌨️ Inline Completions — Real-time ghost text
💬 Chat Responses — Conversational answers
🔧 Agent Actions — Multi-step task execution
📝 Code Edits — Targeted modifications
🏗️ Terminal Commands — Shell command generation

💡 Key Insight: Your code is never stored or used for training (all plans). Prompts are not retained after the response is delivered.

2. Chat Experience

The conversational AI panel in VS Code. Use slash commands, context variables (#), and participants (@) to give Copilot precise context.

Anatomy of a Copilot Chat Prompt

/explain                    ← Slash Command (what to do)
#file:auth.ts               ← Context Variable (what to include)
"How does the login flow work?"  ← Your natural language question

Slash Commands

Command	Action	Command	Action
`/explain`	Explain selected code	`/fix`	Fix problems in code
`/tests`	Generate unit tests	`/new`	Scaffold new project
`/clear`	New chat session	`/help`	Copilot quick reference
`/init`	Generate instructions	`/search`	Workspace search
`/delegate`	Send to coding agent CLI	`/compact`	Compress context CLI

Context Variables (`#`)

Variable	Purpose	Variable	Purpose
`#file`	Include file content	`#selection`	Selected text
`#codebase`	Full workspace context	`#problems`	Error/warning diagnostics
`#changes`	Git changes (diff)	`#fetch`	Fetch a web page
`#terminalLastCommand`	Last terminal output	`#block` / `#class` / `#function`	Code scope

Chat Participants (`@`)

Participant	Purpose
`@workspace`	Project structure & code
`@vscode`	VS Code commands & features
`@terminal`	Terminal shell context
`@github`	GitHub-specific skills
`@azure`	Azure services help

Voice Chat

Install ms-vscode.vscode-speech
Dictate prompts hands-free

💡 Pro Tip: Use /explain + #file to onboard to unfamiliar codebase: "Explain the auth flow in #file:auth.ts"

3. Code Completions

Real-time ghost text suggestions as you type. Copilot predicts your next lines of code based on context.

How Code Completions Work

⌨️ You Type — Code, comment, function signature, or variable name
👻 Ghost Text Appears — Greyed-out inline suggestion (multiple alternatives available)
✅ Accept — Tab — Inserts full suggestion
❌ Dismiss — Esc — Ghost text disappears
🔄 Cycle — Alt+] / Alt+[ — Browse alternative suggestions
✂️ Partial — Ctrl+→ — Accept word by word

Keyboard Shortcuts

Action	Shortcut
Accept full suggestion	`Tab`
Dismiss suggestion	`Esc`
Cycle alternatives	`Alt+]` / `Alt+[`
Accept word by word	`Ctrl+→`

Enable / Disable

Toggle from Copilot icon in status bar
Disable for specific languages in settings
Snooze — temporarily pause suggestions

Completions Model

Change model in settings for inline suggestions
Premium models consume premium requests
Free plan: 2,000 completions/month

💡 Pro Tip: Write a descriptive comment first — Copilot uses it as a prompt to generate the function body below.

4. Next Edit Suggestions (NES)

Predicts your next edit based on recent changes, not just cursor position. Great for repetitive multi-location refactors.

How NES Works

✏️ You Make an Edit — Rename, refactor, add code in one location
🧠 NES Detects Pattern — Analyzes your change and finds similar locations
➡️ Arrow Appears — Indicator points to the next suggested edit location
✅ Tab → Tab — Tab navigates to edit, Tab again accepts it
🔁 Repeat — Tab again for next location

Example: Rename Across Files

// BEFORE (your edit)
const userId = getUser();
console.log(userId);     // ← still old name
fetchData(userId);       // ← still old name

// AFTER (NES applies)
const accountId = getUser();
console.log(accountId);  // ✅ fixed
fetchData(accountId);    // ✅ fixed

Why Use NES

Rename variables across multiple locations
Apply consistent pattern changes
Add missing imports after using a new API

💡 Pro Tip: NES monitors your editing patterns across files and suggests edits at locations you haven't navigated to yet.

5. Inline Chat

Chat with Copilot directly inside your editor or terminal without switching to the sidebar.

In Editor — `⌘I` / `Ctrl+I`

🖱️ Select Code (optional — or place cursor)
💬 ⌘I → Type Prompt — e.g., "Extract into a custom hook"
📊 Preview Diff — Review changes inline
✅ Accept / Discard — Apply edits or reject

In Terminal — `⌘I`

# ⌘I → "Find all .log files > 100MB and delete them"
# Copilot generates:
find . -name "*.log" -size +100M -delete
# ↑ Run or Copy

Vision Input — Attach Images

🖼️ UI mockup → ⌘I + 📎 image → "Build this login form using React + Tailwind"
Supports image-to-code conversion

Key Features

Select code first for targeted edits
Supports context variables & model selection
Preview diff before accepting
Generates and explains terminal commands

💡 Pro Tip: Use inline chat for targeted in-place edits — select a function, ⌘I → "Extract this into a custom hook"

6. Model Selection & Premium Requests

Choose from 20+ AI models. Each model has a premium request multiplier — manage your monthly budget accordingly.

Model Cost Tiers

Tier	Cost	Models	Best For
🆓 Included	0×	GPT-4.1, GPT-4o, GPT-5 mini, Raptor mini	Boilerplate, simple edits
💰 Budget	0.25–0.33×	Grok Code Fast 1, Claude Haiku 4.5, Gemini 3 Flash	Fast iteration, tests
⚡ Standard	1×	Sonnet 4/4.5/4.6, Gemini 2.5/3/3.1 Pro, GPT-5.1/5.2/5.4	Complex refactors, agents
🔥 Premium	3–30×	Claude Opus 4.5/4.6, Opus 4.6 fast mode	High-complexity tasks

💡 Pro Tip: Use included models for routine tasks. Reserve premium models for complex architectural decisions or when you need the highest quality output.

7. Custom Instructions

Markdown files that automatically inject your coding standards, conventions, and project context into every Copilot request.

Instruction Priority (High → Low)

👤 Personal — ~/.copilot/copilot-instructions.md
📁 Path-Specific — .github/instructions/*.instructions.md
🏢 Repository — .github/copilot-instructions.md
🤖 Agent — AGENTS.md
🏛️ Organization — Org-wide rules

Always-On Files

File	Location	Purpose
`copilot-instructions.md`	Root	Repo-wide coding standards
`AGENTS.md`	Root or subfolders	Agent definitions
`CLAUDE.md`	Root, `.claude/`, `~/`	Claude-specific instructions

Task-Specific Instructions

Task	File
Code review	`reviewSelection.instructions`
Commits	`commitMessageGeneration.instructions`
PR descriptions	`pullRequestDescription.instructions`

Generate Instructions

/init — Auto-generate for workspace
Chat debug view → verify loaded files

💡 Pro Tip: Add "Always use single quotes and 2-space indentation" to copilot-instructions.md to enforce consistent style across all Copilot suggestions.

8. Instructions.md Files

Targeted instruction files (.instructions.md) that apply conditionally based on file glob patterns or semantic matching.

Conditional Instructions with `applyTo`

---
name: 'Python Standards'
description: 'Python conventions'
applyTo: '**/*.py'
---
# Python coding standards
- Follow PEP 8 style guide
- Use type hints for all functions

How It Works

When editing src/utils/auth.py:

📝 You're editing: src/utils/auth.py
💬 Copilot Prompt (enriched)
System instructions injected:
✅ python.instructions.md      ← matches **/*.py
❌ react.instructions.md      ← skipped
❌ tests.instructions.md      ← skipped
❌ api.instructions.md        ← skipped

Locations

Scope	Path
Workspace	`.github/instructions/`
Claude fmt	`.claude/rules/`
User	`~/.copilot/instructions/`

💡 Pro Tip: Type /instructions in chat to open the Configure menu and manage your instruction files.

9. Reusable Prompt Files

Saved prompt templates (.prompt.md) you invoke as slash commands. Encode frequent tasks like scaffolding components or running reviews.

How Prompt Files Work

📝 Create .prompt.md
💬 Type /name in chat to invoke
⚡ AI Executes with tools & model

Prompt File Anatomy

---
description: 'Create React form'
agent: agent
tools: ['editFiles', 'search']
model: GPT-5.2
---
Generate a React form component
with validation for ${input:fields}

Locations

Scope	Path
Workspace	`.github/prompts/`
User	Profile `prompts/` folder

Quick Commands

/create-prompt — AI-generate a prompt file
/prompts — Configure prompt files

💡 Pro Tip: Create a /create-api prompt that generates a REST endpoint with controller, service, tests, and OpenAPI spec in one command.

10. Chat Modes & Custom Agents

Define specialized AI agents with scoped tools, models, and handoff workflows for complex multi-step tasks.

Agent Frontmatter Fields

Field	Description
`description`	Shown as placeholder text
`name`	Display name (defaults to filename)
`tools`	List of available tools; omit for all
`agents`	Allowed subagents (`*` = all, `[]` = none)
`model`	AI model (string or prioritized array)
`handoffs`	Sequential workflow transitions
`target`	`vscode` or `github-copilot`
`mcp-servers`	MCP server configs scoped to this agent

Example: Security Review Agent

---
description: 'Security review agent'
tools: ['search', 'readFile']
model: Claude Sonnet 4.6
agents: ['*']
handoffs:
  - label: Fix issues
    agent: implementation
    prompt: Fix the issues above.
    send: false
---
Review code for OWASP Top 10 vulnerabilities...

Where Custom Agents Work

✅ VS Code
✅ JetBrains
✅ Eclipse
✅ Xcode
✅ GitHub.com
✅ Copilot CLI
✅ Coding Agent
✅ Background & Cloud Agents

Locations

Scope	Path
Workspace	`.github/agents/`
Claude fmt	`.claude/agents/`
User	`~/.copilot/agents/`
Org/Enterprise	`.github-private` repo → `agents/`

💡 Pro Tip: Type /agents to configure or /create-agent to generate with AI. Create org-wide agents in .github-private repo to share across all repos.

11. Skills (Agent Superpowers)

Portable folders of instructions, scripts & resources that Copilot auto-loads when relevant. Unlike instructions (guidelines), skills teach capabilities.

How Skills Load — Progressive Context

Request 1: 📋 Metadata

name + description × all skills
AI matches prompt to skill descriptions
✅ Best match selected for loading

Request 2: 📄 SKILL.md

Full instructions loaded into context
Agent reads: Workflow steps, Patterns & rules

Request 3+: 📦 Resources

Scripts, templates, fixtures loaded
Agent uses helpers to complete task

SKILL.md Format

---
name: webapp-testing
description: 'Run and debug web app integration tests'
---
# Testing Workflow
- Use describe + it + AAA pattern
- Use factory mocks for fixtures

Key Features

Invoke via /skill-name in chat
/create-skill — AI-generate a skill
Works across VS Code, CLI & coding agent
Open standard: agentskills.io

Locations

Scope	Path
Project	`.github/skills/<name>/`
Personal	`~/.copilot/skills/<name>/`

💡 Pro Tip: Create a testing skill with SKILL.md + test templates + fixture scripts. Copilot auto-loads it when you ask "help me test".

12. MCP — Model Context Protocol

An open standard that connects Copilot to external tools & services (databases, APIs, browsers). MCP servers expose tools, data resources, and prompt templates that the AI can use during conversations.

MCP Architecture

🧠 Copilot Agent
    ↓ Discovers & calls tools via MCP protocol
🔌 MCP Server (stdio / HTTP transport)
    ↓ Exposes capabilities
🗄️ Databases · 🌐 APIs · 🧭 Browsers · 📁 File Systems · 🐙 GitHub

What MCP Enables

🔍 Database Queries — Query PostgreSQL, MongoDB, etc. directly from chat
🌐 API Integration — Call REST/GraphQL APIs with natural language
🧭 Browser Automation — Scrape web pages, interact with DOM
📁 File Operations — Advanced file system operations
🐙 GitHub Integration — Deep GitHub API integration

Getting Started with MCP

Find or create an MCP server for your tool
Configure it in your workspace or user settings
Copilot automatically discovers and uses available tools

💡 Pro Tip: MCP servers turn Copilot from a code generator into a full-stack assistant that can interact with your entire development environment.

Quick Reference: Essential Shortcuts

Feature	Shortcut	Purpose
Inline Chat	`⌘I` / `Ctrl+I`	Chat inside editor/terminal
Panel Chat	`⌃⌘I` / `Ctrl+Alt+I`	Open sidebar chat
Accept Completion	`Tab`	Insert ghost text
Dismiss Completion	`Esc`	Remove ghost text
Cycle Alternatives	`Alt+]` / `Alt+[`	Browse suggestions
Accept Word	`Ctrl+→`	Partial acceptance
Voice Chat	🎙️	Dictate prompts

Final Thoughts

GitHub Copilot is no longer just an autocomplete tool — it's a comprehensive AI coding platform. The key to getting the most out of it is:

🎯 Be Specific — Use context variables (#file, #codebase) to give Copilot the right context
📝 Customize — Invest time in copilot-instructions.md and prompt files for your workflow
🤖 Automate — Build agents and skills for repetitive tasks
🔌 Extend — Use MCP servers to connect Copilot to your entire toolchain

The developers who master these workflows will spend less time writing boilerplate and more time solving meaningful problems.

"The best Copilot users don't just accept suggestions — they architect the context that produces the right suggestions."

Happy coding! 🚀

Why This Cheatsheet?

1. Copilot Architecture & Flow

How Your Prompt Flows Through Copilot

Context Sources

Output Types

2. Chat Experience

Anatomy of a Copilot Chat Prompt

Slash Commands

Context Variables (#)

Chat Participants (@)

Voice Chat

3. Code Completions

How Code Completions Work

Keyboard Shortcuts

Enable / Disable

Completions Model

4. Next Edit Suggestions (NES)

How NES Works

Example: Rename Across Files

Why Use NES

5. Inline Chat

In Editor — ⌘I / Ctrl+I

In Terminal — ⌘I

Vision Input — Attach Images

Key Features

6. Model Selection & Premium Requests

Model Cost Tiers

7. Custom Instructions

Instruction Priority (High → Low)

Always-On Files

Task-Specific Instructions

Generate Instructions

8. Instructions.md Files

Conditional Instructions with applyTo

How It Works

Locations

9. Reusable Prompt Files

How Prompt Files Work

Prompt File Anatomy

Locations

Quick Commands

10. Chat Modes & Custom Agents

Agent Frontmatter Fields

Example: Security Review Agent

Where Custom Agents Work

Locations

11. Skills (Agent Superpowers)

How Skills Load — Progressive Context

SKILL.md Format

Key Features

Locations

12. MCP — Model Context Protocol

MCP Architecture

What MCP Enables

Getting Started with MCP

Quick Reference: Essential Shortcuts

Final Thoughts

Context Variables (`#`)

Chat Participants (`@`)

In Editor — `⌘I` / `Ctrl+I`

In Terminal — `⌘I`

Conditional Instructions with `applyTo`