GitHub Copilot Certification Fact Sheet¶
Exam Overview¶
Exam Name: GitHub Copilot Duration: 90 minutes Questions: ~65 multiple choice and multiple response Passing Score: 70% Cost: $99 USD Delivery: Online proctored (PSI) Prerequisites: None (Git and IDE familiarity recommended) Validity: 2 years
Official Exam Page - Registration and details GitHub Copilot Documentation - Product docs Copilot Trust Center - Privacy and security
Target Audience¶
- Software developers using AI-assisted coding tools
- Team leads evaluating Copilot for adoption
- DevOps and platform engineers integrating Copilot into workflows
- Administrators responsible for Copilot Business or Enterprise rollout
- Technical consultants advising clients on AI tooling
Domain 1: Introduction to GitHub Copilot (20%)¶
What Copilot Is¶
GitHub Copilot is an AI pair programmer that provides code completions, chat-based assistance, pull request support, and CLI help. It is built on large language models hosted by GitHub on Azure, routed through a GitHub-owned proxy service.
Models in use (subject to change): - OpenAI GPT-4o, o1 family, Codex derivatives for completions - Anthropic Claude 3.5 Sonnet (selectable in chat for eligible customers) - Google Gemini (selectable where available)
About Copilot - Product introduction Copilot Features - Feature map
Copilot Product Family¶
| Product | Audience | Key Features |
|---|---|---|
| Copilot Free | Individual users, trial | Limited completions and chat per month |
| Copilot Pro (Individual) | Individual paid users | Unlimited completions, chat, model choice |
| Copilot Pro+ | Power users | Higher premium request limits, all models |
| Copilot Business | Organizations | Policy controls, content exclusions, audit logs, IP indemnity |
| Copilot Enterprise | Large orgs on Enterprise Cloud | Everything in Business plus knowledge bases, PR summaries, custom models |
Copilot Plans - Plan comparison
Supported Languages and IDEs¶
- Best quality on languages with large public training corpora: Python, JavaScript, TypeScript, Go, Ruby, Java, C#, C, C++, Rust, PHP
- First-class IDE support: VS Code, Visual Studio, JetBrains family, Neovim, Xcode, Eclipse, Azure Data Studio
- Surfaces: IDE inline, IDE chat, GitHub.com chat, GitHub Mobile, CLI (gh copilot), PR review
Domain 2: Responsible Use of GitHub Copilot (15%)¶
Limitations to Communicate¶
- Suggestions can be incorrect, outdated, insecure, or biased
- Training data has a cutoff; very new APIs may be missing or wrong
- Copilot does not know private code unless it is part of the current context (or an Enterprise knowledge base)
- Suggestions can match public code; the duplicate detection filter mitigates this
Responsible Use - Official guidance
Duplicate Detection Filter¶
- Blocks suggestions that match ~150 characters of public code on GitHub
- Configurable at org/enterprise (Business and Enterprise plans)
- Required to be enabled for IP indemnification to apply
- Available to Individual users as a personal setting
Content Exclusions¶
- Path-based rules that prevent files from being used as context and suppress suggestions inside excluded files
- Scoped at repository, organization, or enterprise level
- Applied to both completions and chat
- Do not apply to public code or code already submitted to Copilot
Content Exclusions - Exclusion setup
Data Handling¶
- Prompts, suggestions, and telemetry behavior depends on plan and settings
- Business and Enterprise: prompts and suggestions are not retained or used to train models
- Individual: user can opt in or out of training on prompts
- Data is encrypted in transit; the proxy strips prompts after response in Business/Enterprise
Domain 3: Copilot Plans and Features (15%)¶
Feature Matrix¶
| Feature | Free | Pro | Business | Enterprise |
|---|---|---|---|---|
| Code completions | Limited | Unlimited | Unlimited | Unlimited |
| Chat in IDE | Limited | Yes | Yes | Yes |
| Chat on GitHub.com | Limited | Yes | Yes | Yes |
| Model selection | No | Yes | Yes | Yes |
| Duplicate detection policy | User | User | Admin | Admin |
| Content exclusions | No | No | Yes | Yes |
| IP indemnification | No | No | Yes | Yes |
| Audit logs | No | No | Yes | Yes |
| PR summaries | No | No | No | Yes |
| Knowledge bases | No | No | No | Yes |
| Custom models (preview) | No | No | No | Yes |
Plans for Copilot - Detailed comparison
Billing and Seats¶
- Seats are assigned per user by org or enterprise admins
- Unassigning mid-cycle may or may not refund depending on billing cycle
- Seat utilization can be tracked via API and org insights
- Enterprise accounts can enable Copilot for specific orgs or all orgs
Domain 4: Using GitHub Copilot (25%)¶
Inline Completions (Ghost Text)¶
| Action | Shortcut (VS Code) |
|---|---|
| Accept full suggestion | Tab |
| Accept next word | Ctrl+Right (Cmd+Right) |
| Next suggestion | Alt+] |
| Previous suggestion | Alt+[ |
| Dismiss | Esc |
| Open suggestions panel | Ctrl+Enter |
Copilot Chat Surfaces¶
- Inline chat - Invoked at the cursor (Ctrl+I)
- Chat view/panel - Persistent side panel
- Quick chat - Floating input for quick questions
- Chat on GitHub.com - Repo-aware chat on github.com
- Chat in Mobile - Conversational on iOS and Android
Slash Commands (Chat)¶
| Command | Purpose |
|---|---|
/explain | Explain selected code or concept |
/fix | Propose a fix for a problem |
/tests | Generate tests for selection |
/doc | Generate documentation comments |
/new | Scaffold a new project or workspace |
/clear | Clear chat context |
/help | List chat commands |
Chat Cheat Sheet - Chat reference
Agents and Context Variables¶
| Agent / Variable | Purpose |
|---|---|
@workspace | Whole repo context (VS Code) |
@vscode | Questions about VS Code itself |
@terminal | Shell/terminal help |
@github | GitHub-hosted data (issues, PRs, repo) |
#file | Attach a file as context |
#selection | Use current selection |
#codebase | Broad codebase search |
#terminalLastCommand | Reference last terminal command |
Copilot in the CLI¶
- Install:
gh extension install github/gh-copilot gh copilot suggest "how to tar a directory"- Suggests shell, git, or gh commandsgh copilot explain "grep -rni error ."- Explains a command
Pull Requests¶
- PR summaries (Enterprise) - Auto-generated PR descriptions
- Code review - Copilot can be requested as a reviewer on PRs in eligible plans
- Commit message generation - From IDE source control
Domain 5: Prompt Engineering for Copilot (10%)¶
The 4 S's¶
- Single - One intent per prompt
- Specific - Clear names, types, signatures, inputs, outputs
- Short - Concise prompts; refine iteratively
- Surround - Use neighboring code, open tabs, and comments as context
Prompt Patterns¶
- Zero-shot - Ask once with clear intent
- One-shot - Include one example of the desired output
- Few-shot - Provide several examples
- Chain prompting - Build up context with follow-ups in chat
Practical Techniques¶
- Write a descriptive function signature and docstring before asking for the body
- Keep related files open in tabs so Copilot can draw context
- Use types, schemas, and imports to anchor the prompt
- For refactors, explain the current behavior and the desired behavior
Prompt Engineering for Copilot - Official guide
Domain 6: Developer Use Cases (15%)¶
Core Use Cases¶
- New code generation from comments and signatures
- Test generation (unit, integration, property)
- Refactoring and rename suggestions
- Documentation (README, inline, API)
- Learning unfamiliar frameworks (ask Copilot to explain)
- Code review assistance and PR summaries
- DevOps: Dockerfile, GitHub Actions YAML, Terraform, Bash
Regulated and Sensitive Scenarios¶
- Use content exclusions for secrets, license-sensitive code, regulated data
- Enable duplicate detection to reduce public code matches
- Turn off Copilot in repos where prohibited by policy via org settings
Exam Tips¶
High-Priority Topics (by weight)¶
- Using Copilot (25%) - slash commands, agents, context, keyboard shortcuts
- Introduction to Copilot (20%) - plans, models, features, surfaces
- Responsible Use (15%) - duplicate detection, content exclusions, IP indemnity
- Plans and Features (15%) - feature matrix per plan
- Developer Use Cases (15%) - real-world patterns
- Prompt Engineering (10%) - 4 S's, zero/one/few-shot
Key Differentiators to Remember¶
- Duplicate detection blocks matches to public code; content exclusions block files from context
- IP indemnity requires Business/Enterprise plus duplicate detection enabled
- Knowledge bases and PR summaries are Enterprise-only
@workspaceis VS Code;@githubtalks to GitHub.com data- Individual plan users can opt their prompts into training; Business/Enterprise prompts are not used for training