In a guide aimed at beginners, GitHub presents a way to review changes created by the coding agent within the GitHub Copilot app, without switching between the code editor, terminal window, and web browser. The app brings these tasks together in adjacent panels that help developers inspect the change, run the project, and preview the result before accepting it.
The guide’s central idea is not to accept code produced by the agent immediately, but to complete the review, execution, and testing cycle within a single session. According to the article, after confirming the result, users can accept the change directly from the app and create a pull request.
Reviewing Changes Through the Diff Panel
The diff panel displays a comparison between the state of the files before and after the modification, highlighting added lines in green and deleted lines in red. This allows users to understand the scope of the change instead of relying only on the final result suggested by the agent.
GitHub notes that users can accept changes, leave comments on them, or ask Copilot to make additional modifications. Final approval of what will enter the project remains in the developer’s hands.
Running the Project from the Terminal Panel
Reviewing the code is not enough; the terminal panel allows users to run commands within a Copilot app session and read their output. The guide presents an example of a web project in which a development server script can be added to open the client folder and then run the npm run dev command.
After setting up the script, the user presses the run button to start the site server. The app also allows more than one terminal window to be open at the same time, enabling users to switch between different commands and sessions while examining the project.
Previewing and Modifying the Interface
For projects that include a user interface, the browser panel provides a way to open the application and test the new feature as users see it. This section includes the Pick & Polish tool, which allows users to select an element in the interface and then ask the agent to modify it.
After making the changes, the development server script can be run again to preview what changed and verify that the fix or improvement appears in the actual application.
What Does This Mean for Developers?
These features connect three stages that are often separate when working with code generated by artificial intelligence: inspecting changes, running the project, and testing its behavior or interface. Bringing them together in one place may reduce the loss of context caused by switching between windows, but it does not eliminate the need for human review or actual testing.
GitHub suggests a short checklist before accepting any automatically generated change: What changed? Does the code work? And does it actually perform the required function? These questions do not by themselves prove that the code is correct, but they provide a practical checkpoint before integrating the changes.
The practical limitation shown by the source is that the guide explains the workflow specifically within the GitHub Copilot app and does not provide a comparison with other tools or independent measurements of code accuracy or test quality. Therefore, the value of the panels remains tied to how well the app fits the developer’s workflow and the developer’s ability to inspect the results, not merely to the presence of the agent.