GH-600 · Domain 1: Prepare agent architecture and SDLC processes · 18% of the exam
Task 1.1: Integrate agents into the software development lifecycle (SDLC)
Choosing which steps of delivery an agent should take, spotting and avoiding agent anti-patterns, and defining an agent task's inputs, outputs and success criteria.
Study it
Which delivery steps to give an agent, and the tasks to keep for people
Try each one before opening the answer. Every option is explained, with the GitHub documentation page that proves it.
Question 1 · choose 1
A team lead is choosing one backlog item to hand to Copilot cloud agent in the first week of a pilot, so that reviewers can learn how it works on a low-risk change. Which item fits that goal best?
AAdd unit tests for the date-parsing helpers, naming the test framework and the folder they belong in
BFind the cause of the checkout outage that started in production an hour ago
CRedesign how three services share authentication, keeping the change consistent across all of their repositories
DExplore whether the team should adopt event sourcing, with no requirements agreed yet
Show the answer and why
AAdd unit tests for the date-parsing helpers, naming the test framework and the folder they belong in
Correct
Well-defined work such as improving test coverage is among the tasks GitHub suggests starting with.
BFind the cause of the checkout outage that started in production an hour ago
Incorrect
Production-critical issues and incident response are tasks GitHub suggests keeping for people.
CRedesign how three services share authentication, keeping the change consistent across all of their repositories
Incorrect
Broad, cross-repository work that needs design consistency and touches authentication is better done by developers.
DExplore whether the team should adopt event sourcing, with no requirements agreed yet
Incorrect
Ambiguous, open-ended tasks without a clear definition are better worked through by people.
GitHub suggests starting Copilot cloud agent on simpler, well-scoped tasks such as fixing bugs, improving test coverage, updating documentation or addressing technical debt. Production-critical issues, incident response, security and authentication work, broad cross-repository refactors and ambiguous tasks are ones you may choose to work on yourself.
A maintainer wants a Copilot automation to re-check a pull request's changelog entry every time someone on the team pushes new commits to that pull request, not only when it is first opened. Which trigger should the automation use?
AWhen a pull request is opened, filtered to pull requests that touch CHANGELOG.md
BOn an hourly schedule that scans every open pull request
CWhen an issue is created, with a search query that matches the pull request
DWhen a pull request is synchronized, with an optional files-changed filter
Show the answer and why
AWhen a pull request is opened, filtered to pull requests that touch CHANGELOG.md
Incorrect
The opened trigger runs once, when the pull request is created; a filter doesn't make it run on later pushes.
BOn an hourly schedule that scans every open pull request
Incorrect
A schedule runs at fixed intervals whether or not anything was pushed, so it doesn't follow each push.
CWhen an issue is created, with a search query that matches the pull request
Incorrect
This trigger runs when an issue is opened; it doesn't react to pushes to a pull request.
DWhen a pull request is synchronized, with an optional files-changed filter
Correct
This trigger runs each time new commits are pushed to a pull request in the repository.
Automations can run on a schedule (hourly, daily or weekly), when an issue is created, when a pull request is opened, or when a pull request is synchronized, which means new commits are pushed to it. The pull request triggers accept a search query filter and a filter on the files changed.
A reviewer posts six separate comments, one after another, each mentioning @copilot, on a pull request that Copilot cloud agent opened. Copilot starts after the first comment and later pushes commits that address the comments piecemeal and partly undo each other. What should the reviewer do next time?
AHave a teammate without write access post the comments, so Copilot only reads them at the end
BPost the comments on the linked issue instead, where Copilot collects them for its next push
CWrite the comments as one review with Start a review, then submit them all at once
DMention @copilot in the pull request description instead of in review comments
Show the answer and why
AHave a teammate without write access post the comments, so Copilot only reads them at the end
Incorrect
Copilot only responds to comments from people with write access; comments from others are not acted on at all.
BPost the comments on the linked issue instead, where Copilot collects them for its next push
Incorrect
Copilot reads the issue when it is assigned; follow-up feedback belongs on the pull request, batched in one review.
CWrite the comments as one review with Start a review, then submit them all at once
Correct
Copilot starts looking at comments as soon as they are submitted, so a single review makes it work on the whole set together.
DMention @copilot in the pull request description instead of in review comments
Incorrect
Copilot keeps the title and body up to date itself; the documented way to send feedback is through comments, batched in one review.
Working on a Copilot pull request is like working with a developer: you mention @copilot in comments or push commits yourself. Because Copilot starts looking at comments as soon as they are submitted, GitHub recommends batching several comments with Start a review rather than Add single comment, so Copilot works on the entire review. Only comments from users with write access are acted on.
When creating a new repository on GitHub.com, a developer types a description of a small Rust command-line tool into the Prompt field and clicks Create repository. What does Copilot produce?
AA draft pull request with scaffolded starter code
BA plan in Copilot Chat that waits for approval
CAn issue assigned to Copilot that lists the files to create
DA set of files committed directly to the default branch
Show the answer and why
AA draft pull request with scaffolded starter code
Correct
Seeding a repository creates a draft pull request with the scaffolded code.
BA plan in Copilot Chat that waits for approval
Incorrect
Seeding goes straight to scaffolded code in a pull request; no plan is shown first.
CAn issue assigned to Copilot that lists the files to create
Incorrect
No issue is created; Copilot opens a draft pull request.
DA set of files committed directly to the default branch
Incorrect
The code arrives as a draft pull request for review, not as a direct commit.
You can start a Copilot cloud agent task by assigning an issue, by starting with a prompt, or by seeding a new repository. When you describe what to build in the Prompt field while creating a repository, Copilot opens a draft pull request with scaffolded code.
A maintainer of a public open source repository wants Copilot to label new issues automatically, but the Automations pane is not offered in the repository. What explains this?
AAutomations require the Copilot Enterprise plan
BOnly repository administrators can create automations
CAutomations must first be switched on in the repository's copilot-setup-steps.yml file
DAutomations are available only in private and internal repositories
Show the answer and why
AAutomations require the Copilot Enterprise plan
Incorrect
Automations are available with Copilot Pro, Pro+, Max, Business and Enterprise.
BOnly repository administrators can create automations
Incorrect
Any user with write access to a repository can create automations in it.
CAutomations must first be switched on in the repository's copilot-setup-steps.yml file
Incorrect
That file sets up the agent's environment; it doesn't control whether automations are offered.
DAutomations are available only in private and internal repositories
Correct
Automations are not available in public repositories.
For automations to be available, the repository must be private or internal, Copilot cloud agent must be enabled for it, and the organization must allow both cloud agent and automations. Automations come with the Pro, Pro+, Max, Business and Enterprise plans, and anyone with write access can create them.
The first time a developer asks Copilot cloud agent to create a pull request in a repository, what does Copilot offer to help it work better there in future?
AA wizard that writes a copilot-setup-steps.yml file
BA checklist of acceptance criteria to add to the issue
CA comment with a link to generate custom instructions automatically
DAn option to copy the custom instructions from another repository in the organization
Show the answer and why
AA wizard that writes a copilot-setup-steps.yml file
Incorrect
No wizard is offered; you write the setup steps file yourself.
BA checklist of acceptance criteria to add to the issue
Incorrect
Copilot doesn't add a checklist to the issue.
CA comment with a link to generate custom instructions automatically
Correct
Copilot leaves a comment with a link to automatically generate custom instructions.
DAn option to copy the custom instructions from another repository in the organization
Incorrect
The offer is to generate instructions for this repository, not to copy them.
Repository custom instructions tell Copilot how to understand the project and work in it. The first time you ask Copilot to create a pull request in a repository, it leaves a comment with a link to generate custom instructions automatically; you can also ask it to generate them at any time.
A designer drags a 5 MiB PNG mockup into an issue that is then assigned to Copilot cloud agent, but the agent's work shows no sign of the mockup. What is the likely reason?
ACloud agent reads only text, so images need Copilot Chat
BImages larger than 3.00 MiB are removed from the request
CPNG files aren't supported, only JPEG screenshots
DImages must be committed to the repository and linked by path
Show the answer and why
ACloud agent reads only text, so images need Copilot Chat
Incorrect
Cloud agent processes images such as screenshots and designs as part of the task.
BImages larger than 3.00 MiB are removed from the request
Correct
The maximum image size cloud agent accepts is 3.00 MiB.
CPNG files aren't supported, only JPEG screenshots
Incorrect
PNG, JPEG, GIF and WebP images are supported.
DImages must be committed to the repository and linked by path
Incorrect
You can drag an image into an issue body or a prompt.
Cloud agent processes images, such as screenshots, sketches or designs, that you add to an issue body or a prompt; PNG, JPEG, GIF and WebP files are supported. The maximum image size is 3.00 MiB, and larger images are removed from the request.