GitHub Copilot app for Beginners: Write your first prompt

GitHub’s new guide explains that a prompt is simply a description of the desired outcome, not a polished query. Users can begin by describing a small change in plain language, such as adding a new sorting option, and let the tool refine the request through iteration. The guide emphasizes that the first attempt does not need to be perfect, as additional details can be provided later to adjust the output. This approach lowers the barrier to entry for new users unfamiliar with AI-powered coding assistance.

To start a task, users must connect the GitHub Copilot app to a project, either a GitHub repository or a local folder. This connection provides the necessary context for the AI to analyze the codebase and execute the requested change. The guide notes that users can select a previously worked project or add a new one directly from the app’s home screen. For code stored only on a local machine, a local folder can be added instead, ensuring the session has access to the relevant files.

The guide also highlights the option to choose between different AI models within the app, each with varying strengths for tasks like complex reasoning or quick execution. Beginners are advised to start with the default model, as understanding model differences is not required to begin. Users can switch models later if a task demands more advanced reasoning or if the initial results are unsatisfactory, treating the model selector as an optional tool rather than a mandatory setup step.

In addition to text prompts, the GitHub Copilot app supports voice input, converting spoken descriptions into editable text for the prompt box. This feature allows users to think aloud and refine their requests before finalizing them. The guide also mentions options to change the session agent or enable remote control, enabling users to switch devices without losing progress. These features are presented as optional tools for users who need greater flexibility or control over their tasks.