Explore / GitHub Copilot (GH-300) Exam Prep / Practice questions
GitHub Copilot (GH-300) practice questions
10 questions from the bank behind this course's timed mock. Free to read and to answer, no account needed, and each has its answer and an explanation. They belong to GitHub Copilot (GH-300) Exam Prep.
Answer them one at a time →Read lesson 1 free
Question 1
A candidate is booking GH-300 and wants to plan time per question. They search the official study guide for the exam length and cannot find it. What is the BEST way to confirm how long the exam lasts?
- Use the duration quoted in a popular exam-prep blog post from last year
- Check the official GH-300 exam page on Microsoft Learn shortly before booking, because it states the current duration and these details change
- Assume it is the same length as other Microsoft role-based exams and plan around that
- Ask GitHub Copilot Chat how long the GH-300 exam is
Show the answer
Answer: B. Check the official GH-300 exam page on Microsoft Learn shortly before booking, because it states the current duration and these details change
The study guide does not give the length, but the official exam page does (currently 100 minutes), and you should check it again before booking. Old blog posts and AI answers can be out of date or simply wrong, which is the same reason you don't treat AI output as an authority.
Question 2
A front-end developer accepts a Copilot-generated sign-up form. In review, a colleague finds that the form fields have no labels a screen reader can announce, so blind users can't complete sign-up. Which responsible AI principle is most directly at stake?
- Reliability and safety
- Inclusiveness
- Privacy and security
- Transparency
Show the answer
Answer: B. Inclusiveness
Inclusiveness means AI systems should empower everyone and engage people, and an accessible UI is how that shows up in a Copilot workflow. Reliability and safety is about the code working correctly and safely. A form that technically works but shuts out screen-reader users is an inclusiveness failure.
Question 3
A developer signs in to VS Code with their work GitHub account to use their company's Copilot Business plan. The Copilot icon in the Status Bar shows an error, but the same machine worked fine earlier with their personal account. What should they check first?
- Whether the chat.disableAIFeatures setting is turned on for the workspace
- Whether they have used up their monthly allowance of 2,000 completions
- Whether an organisation admin has actually assigned them a Copilot seat
- Whether Node.js 22 or later is installed on the machine
Show the answer
Answer: C. Whether an organisation admin has actually assigned them a Copilot seat
An admin assigns Business and Enterprise seats, so an error on a work account usually means no seat has been assigned. The 2,000-completion limit belongs to Copilot Free, not to a Business seat.
Question 4
A backend developer gives Agent Mode a task. During the run, the agent edits four source files and also runs a command that changes data in a shared staging database. The result is wrong, and the developer restores the checkpoint from before the run. What is the state afterwards?
- The four file edits and the staging database changes are both rolled back
- Nothing is undone, because checkpoints only record the chat history and not file changes
- The four file edits are undone, but the staging database changes remain and must be fixed separately
- The staging database changes are rolled back, but the file edits stay until each one is undone separately
Show the answer
Answer: C. The four file edits are undone, but the staging database changes remain and must be fixed separately
Restoring a checkpoint undoes the agent's file changes, but it cannot undo anything the agent did to external services such as a database. That is why you should read every command before you click Allow. The idea that everything is rolled back is the tempting mistake, because checkpoints only cover the workspace files.
Question 5
A developer requested a Copilot code review on a pull request, then pushed three more commits to fix what a human reviewer found. She wants Copilot to look at the updated code. What should she do?
- Add a new rule to .github/copilot-instructions.md and push it to the base branch
- Use the Copilot icon in the description field and choose "Summary" again
- Click the re-request (sync) icon next to Copilot's name in the Reviewers menu
- Close and reopen the pull request so that Copilot reviews it automatically
Show the answer
Answer: C. Click the re-request (sync) icon next to Copilot's name in the Reviewers menu
You start a new Copilot review on an updated PR with the re-request (sync) icon next to Copilot in the Reviewers menu. Choosing "Summary" only drafts a PR description and doesn't review any code.
Question 6
An organization owner on Copilot Business wants members to use Copilot features on github.com, such as chat and pull request summaries. In the IDE, the team should only get generally available functionality. Which policy configuration meets this goal?
- Enable 'Copilot in GitHub.com' and disable 'Editor preview features'
- Enable 'Copilot in GitHub.com' and block 'Suggestions matching public code'
- Enable 'Copilot code review' and disable 'MCP servers in Copilot'
- Disable 'Copilot in GitHub.com' and enable 'Editor preview features'
Show the answer
Answer: A. Enable 'Copilot in GitHub.com' and disable 'Editor preview features'
'Copilot in GitHub.com' turns on the website features, and turning off 'Editor preview features' keeps preview functionality out of the IDE, so the IDE only gets GA features. Blocking public-code matches is a real policy, but it filters suggestions. It doesn't control preview features.
Question 7
A compliance officer is reviewing GitHub Copilot before a company-wide rollout. Before approving it, they need to know where prompts are screened before any model sees them. Which answer is accurate?
- Only inside the developer's IDE extension, which filters the prompt locally and then sends it straight to the model
- Inside the model itself, which refuses unsafe prompts while it generates code, with no separate screening layer
- On a proxy hosted in a GitHub-owned Microsoft Azure tenant, which screens prompts before they reach the model
- In the company's own GitHub Enterprise Server instance, which must host the filtering service on-premises
Show the answer
Answer: C. On a proxy hosted in a GitHub-owned Microsoft Azure tenant, which screens prompts before they reach the model
The prompt travels over HTTPS to a GitHub-owned proxy in a Microsoft Azure tenant, and that proxy screens it before the model receives it. The IDE gathers context and builds the prompt, but the screening itself happens at the proxy, not on the developer's machine.
Question 8
A backend developer is writing a new function in src/billing/invoice.ts and finds that Copilot's inline suggestions ignore the project's Money helper type. That type is defined in src/billing/money.ts, which is closed. The editor has a dozen unrelated frontend files open from an earlier task. What should the developer do first to improve the inline suggestions?
- Open src/billing/money.ts and close the unrelated frontend files, so neighboring tabs give relevant context
- Type #codebase in the editor above the function so inline suggestions search the whole repository
- Add the Money type rules to .github/copilot-instructions.md and wait for the next suggestion
- Switch to a different model in the Copilot model picker
Show the answer
Answer: A. Open src/billing/money.ts and close the unrelated frontend files, so neighboring tabs give relevant context
Inline suggestions draw on the code around the cursor and on open tabs, so opening the relevant file and closing unrelated ones changes the context directly. #codebase is a Chat #-reference and does nothing in inline suggestions in the editor.
Question 9
A developer writes this comment: 'Orders strictly over 1000 get an extra 2%.' Copilot's inline suggestion uses `if total >= 1000:`. The developer has time to run only one input before committing. Which input value is most likely to expose the defect?
- 0
- 999
- 1000
- 2000
Show the answer
Answer: C. 1000
Only an input of exactly 1000 behaves differently under `>` and `>=`, so the boundary value is the one that exposes the bug. An input of 2000 is over the threshold under both operators, so it passes either way and proves nothing about the boundary.
Question 10
A repository administrator wants Copilot to stop using one file, the `config.json` at the root of the repository, as context. The repository also has several other files named `config.json` in subfolders, and the team still wants Copilot to use those. Which entry should the admin add to the repository's Content exclusion setting?
- "/**/config.json"
- "config*"
- "/config.json"
- "config.json"
Show the answer
Answer: C. "/config.json"
At repository level, a leading slash makes the path relative to the repository root, so `"/config.json"` matches only that one file. A bare `"config.json"` with no slash matches a file of that name anywhere in the repository, so it would also exclude the subfolder copies the team wants to keep.
The full timed mock
50 questions in 100 minutes, with a pass mark of 70%. Each paper is drawn from a bank of questions written for the mock alone, and these 10 are left out of it. Full access opens it, with every level of the course.
LevelUpWith is independent of the organisation that sets this exam. These questions were written for this course; they are not taken from the exam.