2 years. GitHub is moving to Microsoft's recertification process.
Study time, our estimate
20 to 35 hours
GH-300Associate
Quick to prepare for if you use Copilot daily, and most of what you do not know yet is the admin side: content exclusions, policies, audit logs and how data flows. Those parts are worth knowing anyway if you lead a team.
Take it if
You use Copilot at work, or you roll it out across a team.
Skip it if
Your team has standardized on a different coding agent. The exam is specific to Copilot.
Before you start
Check yourself against this list first. If something is missing, close that gap before you pay for an exam.
You code in at least one languageThe exam assumes everyday development.
You know GitHub basicsRepos, branches and pull requests.
You have Copilot accessYou need to try each feature, including the CLI. Learn more
2 years. GitHub is moving to Microsoft's recertification process.
Delivery
Pearson VUE, proctored
Languages
English, Spanish, Portuguese (Brazil), Korean, Japanese
Retake
After 24 hours for the first retake
If you fail
Wait 24 hours after a first attempt
What the exam covers
Use GitHub Copilot responsibly15–20%
Use GitHub Copilot features25–30%
Understand GitHub Copilot data and architecture10–15%
Apply prompt engineering and context crafting10–15%
Improve developer productivity with GitHub Copilot10–15%
Configure privacy, content exclusions, and safeguards10–15%
What to focus on
Every surface: inline suggestions, chat, the Copilot CLI, agent mode, and Copilot Edits.
Agent sessions, sub-agents, and MCP inside the editor.
Spaces, Spark, pull request summaries, and instructions files for review standards.
How a prompt is built, how it passes through the proxy filter, and what happens afterward.
Org-wide work: policies, audit log events, and managing seats through the REST API.
Content exclusions, public code filtering, and what Copilot can and cannot see. GitHub says the CLI and agent mode do not honor exclusions.
Everything on one mapEvery topic and key idea in this certification.
mindmap
root((GH-300))
Responsible use
Validate output
Limits
Harms
Features
Surfaces
Instructions
Org controls
Data and prompting
Prompt flow
Weighting
Fresh context
Productivity and privacy
Good uses
Content exclusions
Exclusion gaps
Public code
Lesson 1 of 4, 15–20% of the exam
Responsible use
Every Copilot suggestion is still your code.
Validate output
Review for correctness and security, as you would any code.
Limits
LLMs can be wrong and sound certain.
Harms
Know the risks and how to reduce them.
The validation loopCopilot drafts. You stay accountable.
flowchart LR
A["Copilot suggestion"] --> B{"Understand it?"}
B -->|No| C["Ask or rewrite"]
B -->|Yes| D{"Correct and secure?"}
D -->|No| C
D -->|Yes| E["Tests and review"]
C --> A
Common trap
Treating passing tests as proof the code is right.
Try it yourself
Ask for a function, then list what you would check before merging.
Inline suggestions, chat, the CLI and agent mode each suit different work.
Instructions
.github/copilot-instructions.md applies repo-wide, and NAME.instructions.md files under .github/instructions target paths.
Org controls
Policies, audit log events, and seat management through the REST API.
Which Copilot surface?Match the size of the job.
flowchart TD
A{"What is the job?"} -->|"Finish this line"| B["Inline suggestion"]
A -->|"Ask a question"| C["Chat"]
A -->|"Work in the terminal"| D["Copilot CLI"]
A -->|"Change many files"| E["Agent mode"]
A -->|"Run in the background"| F["Cloud agent"]
Common trap
Studying only IDE features. The August 2026 update added the Copilot CLI.
Try it yourself
Add repo instructions, then use the CLI and agent mode on the same task.
Context is gathered, the prompt is built, and it passes through a proxy with filtering.
Weighting
GitHub says Copilot gives extra weight to instructions at the start and end of a prompt.
Fresh context
GitHub recommends starting a new session for a new task.
What happens to a promptKnow each stage for the data questions.
sequenceDiagram
participant I as Editor
participant P as Proxy
participant M as Model
I->>I: Gather context and build the prompt
I->>P: Send the prompt
P->>P: Filter
P->>M: Forward
M-->>P: Suggestion
P-->>I: Post-processed result
Common trap
Mixing up zero-shot and few-shot. Few-shot includes examples.
Try it yourself
Rewrite a vague prompt with direct rules and an example, then compare.
Admins are tested on what Copilot can and cannot see.
Good uses
Tests, refactoring, documentation and modernizing old code.
Content exclusions
Set in repository or org settings. Changes can take up to 30 minutes to reach IDEs that already loaded settings.
Exclusion gaps
GitHub states that Copilot CLI and agent mode in IDE chat do not support content exclusion.
Public code
A filter can block suggestions that match public code.
Does content exclusion protect this file?Based on GitHub's documentation. Recheck it, as support changes.
flowchart TD
A{"Which Copilot surface?"} -->|"Inline suggestions and chat"| B["Exclusion applies"]
A -->|"Copilot CLI"| C["Not supported"]
A -->|"Agent mode in IDE chat"| C
C --> D["Use other controls for secrets"]
Common trap
Assuming content exclusion protects a file from every Copilot surface. It does not.
Try it yourself
Set an exclusion in a test repo and check where it applies.
Exam facts checked on October 31, 2026 against GH-300 study guide, skills as of August 7, 2026. Lessons are our own explanations, so use the official docs as the final word, and confirm price and dates before booking.
