• 🌐 GitHub.com UI — Copilot everywhere in the web interface
• 💻 Your IDE — The classic, the OG, the one your coworkers envy
• ⌨️ GitHub CLI — Terminal power users, this one's for you
• 📱 Copilot App — The new standalone experience
• 🚀 Hackathon Strategy — Putting it all together on game day

🤖 Copilot Coding Agent

• Assign a GitHub Issue to @copilot
• It reads the issue, writes code, opens a PR
• You review and merge
• You look incredibly productive

🗺️ Copilot Workspace

• Describe a feature in natural language
• Copilot plans the implementation
• Review and edit the plan
• Generate the code from the plan

• Just start typing — ghost text appears automatically
• Press Tab to accept the whole suggestion
• Press → to accept word-by-word (underrated!)
• Alt+] / Alt+[ to cycle through suggestions
• Open the completions panel to see up to 10 alternatives

# Calculate distance between two GPS coordinates
# using the Haversine formula
def haversine(lat1, lon1, lat2, lon2):
    # ✨ Copilot writes the whole thing...
    R = 6371  # Earth radius in km
    dlat = radians(lat2 - lat1)
    dlon = radians(lon2 - lon1)
    a = sin(dlat/2)**2 + cos(radians(lat1)) \
        * cos(radians(lat2)) * sin(dlon/2)**2
    return R * 2 * asin(sqrt(a))
        

You wrote the comment. Copilot wrote everything else. You're basically a software architect now.

Edit Mode

• Attach multiple files to the context
• Describe the change you want
• Review a unified diff across all files
• Accept or discard changes file by file

Agent Mode 🔥

• Copilot autonomously plans and executes
• Reads files, runs terminal commands
• Iterates until the task is done
• You watch and approve

Core Slash Commands

Model Selection with /model

• 🤖 Auto — intelligently picks the best model, reduces rate limits
• 🧠 Claude Opus 4.5 (default) — complex architecture, deep debugging
• ⚡ Claude Sonnet 4.5 — fast, cost-effective, handles most tasks
• 🔍 GPT-5.2 Codex — code generation, second-opinion reviews

# 1. EXPLORE — understand first, don't code yet
Read the auth files but don't write code yet

# 2. PLAN — create a structured plan
/plan Add OAuth2 with Google and GitHub providers

# Copilot asks clarifying questions,
# creates a checklist plan, saves to plan.md,
# and waits for YOUR approval before proceeding

# 3. IMPLEMENT — execute the approved plan
Proceed with the plan

# 4. VERIFY — don't skip this step
Run the tests and fix any failures

# 5. COMMIT — clean history
Commit these changes with a descriptive message
        

♾️ Infinite Sessions

Never worry about context limits. The CLI automatically compacts conversation history while preserving essential information.

/session           # View session info
/session plan      # See current plan
/context           # Visualize token usage
/compact           # Manually compact (rarely needed)
/clear             # Start fresh (better for unrelated tasks)
        

🗂️ Multi-Repo Work

# Add other repos to the session
/add-dir /path/to/other-service
/list-dirs          # See what's in scope
        

🚀 Autopilot & /fleet

• Switch to autopilot mode for long-running tasks that don't need supervision
• /fleet breaks a task into parallel subtasks run by sub-agents simultaneously
• /delegate sends work to the cloud agent so you can keep coding locally

🛠️ Tool Permissions & Context

# Allow git commands without asking each time
copilot --allow-tool='shell(git:*)'

# Deny a specific dangerous command
copilot --deny-tool='shell(git push)'

# Reset all previously approved tools
/reset-allowed-tools
        

🖼️ Image Input

• Drag & drop images directly into the CLI input
• Paste from clipboard with Ctrl+V
• "Implement this design: @mockup.png"

The "My Work" View

• See all work in motion in one place — agent sessions, issues, PRs, automations
• Each task runs in its own isolated environment with its own branch and state
• Start sessions from an issue, a PR, a code prompt, or a previous session
• Pause and resume sessions across projects without losing context

Agentic Power Features

• 🤖 Run multiple agents in parallel on different tasks simultaneously
• 📋 Automate repeatable workflows — triage, dependency updates, release notes
• 🔍 Review plans and diffs, leave feedback, guide the next iteration — all in-app
• ✅ Agent Merge: let the agent address review comments and auto-merge when conditions are met

Built-in Review Loop

• Review the agent's plan before it codes — course correct early
• See diffs, run validation commands, open live previews — all without leaving the app
• Leave feedback and the agent iterates
• Sessions are tied to PRs — full GitHub review workflow intact

Hackathon Use Cases

• 🏗️ Plan phase: create issues, the app turns them into agent tasks with isolated branches
• ⚡ Build phase: review agent diffs while you focus on the unique logic only you can write
• 🎨 Polish phase: spin up an agent for README, docs, and tests while you prep the demo
• 🚀 Ship phase: Agent Merge handles review comments so you can focus on the pitch

❌ Weak Prompts

• "write a function"
• "fix this"
• "make it better"
• "add auth"

✅ Strong Prompts

• "Write a TypeScript function that takes an array of user objects and returns them sorted by creation date, with null dates last"
• "Fix the race condition in this async function — it should wait for all promises before returning"

Repository Instructions

Create a .github/copilot-instructions.md file:

# Project: HackathonApp

## Stack
- React + TypeScript frontend
- FastAPI Python backend
- PostgreSQL database
- Deployed on Railway

## Conventions
- Use async/await, never callbacks
- All API responses use { data, error }
- Prefer functional components
- Write tests for all API endpoints

## Context
This is a hackathon project for
tracking team productivity metrics.
        

Personal Instructions

Set in GitHub Copilot settings for your personal style:

• Preferred programming language style
• Error handling preferences
• Testing framework of choice
• Code comment verbosity level

Agent Profiles

A Markdown file with YAML frontmatter at .github/agents/NAME.md:

---
name: api-builder
description: Builds REST API endpoints
  following our conventions
---

You are an API specialist for this
project. Always:
- Use FastAPI with async handlers
- Return { data, error } objects
- Add input validation with Pydantic
- Write tests for every endpoint
- Follow our OpenAPI spec in /docs
        

Where Custom Agents Work

• 🌐 GitHub.com — agents tab, issue assignment, PRs
• 💻 VS Code, JetBrains, Eclipse, Xcode
• ⌨️ GitHub CLI

Scope Levels

• .github/agents/ → repo-level (team shared)
• Org's .github repo → org-wide
• Enterprise .github-private → everyone

Where Skills Live

Where to Find Skills

• 🛒 Community-built: github/awesome-copilot
• 🤖 Anthropic's collection: anthropics/skills
• ⌨️ Install via CLI: gh skill install <repo>
• ✍️ Write your own in minutes

🌐 GitHub.com

• ✨ Copilot Chat sidebar
• 📝 PR summary generation
• 🐛 Copilot code review
• 🤖 Assign issues to @copilot
• 🔍 Natural language search
• ⚡ Actions debugging help

💻 IDE (VS Code)

• Tab → accept suggestion
• Alt+] → next suggestion
• Cmd+I → inline chat
• /explain, /fix, /tests, /doc
• #codebase, #file, #terminal
• @workspace, @vscode

🖥️ CLI & App

• copilot → start CLI session
• /plan → structured planning
• /delegate → cloud agent PR
• /fleet → parallel sub-agents
• Desktop app: macOS/Win/Linux
• Agent Merge for auto-shipping

🤖 Skills & Agents

• gh skill install <repo>
• .github/skills/ (project)
• ~/.copilot/skills/ (personal)
• .github/agents/NAME.md
• name/description/prompt
• Assign agents to issues