5 GitHub Copilot features you might not know about

1. Drag and drop your problems

You have written some code, and you have a compilation error. You can see the error in your “problems” window of your IDE but you are not sure what to do about it.

You can right click on the problem line and choose from “Fix using Copilot” or “Explain using Copilot” from the menu.

The feature you might not know about is that you can also click in the middle of this line and drag it into the Copilot chat window. This adds the error as an additional context to your chat, so that it will be considered by the tool in what you then discuss with Copilot.

2. Fixing the “Generating a markdown file in chat” problem

If you have tried to create a markdown file in Copilot Chat, you may well have come across what I think is a bit of a bug…. Rather than seeing the entire source code that you can insert into your file as a single block, the file starts being rendered, making it impossible to copy or insert it.

This bug only occurs in the chat window and it’s annoying when it does. There are two ways I found to overcome this problem (and there may well be more).  The first is that you can make the request from outside of the chat window (i.e. use CTRL+I in the file itself to get the context window).

The second alternative is to use Agent mode… and to understand what that is…

3. Ask, edit or agent?

The default mode to interact with Copilot in the chat window is “ask” – this is designed to simulate a conversation… you can ask Copilot questions and it will answer you. However clicking on the little dropdown lets you change the tool into Edit or Agent mode.

This is a relatively new feature in Copilot – it was only launched in February this year. Edit mode allows you to define tasks for copilot to do, where it will just go and do it directly. If you are in ask mode, Copilot presents you with a potential solution which you can then choose to use, abandon or refine. In edit mode, Copilot just makes the changes directly. You still get the chance to keep or undo the change, but you don’t get the explanations or preview.

This avoids the problems mentioned in the previous item in this article, but it can also be significantly quicker if you are doing routine tasks that would require edits to multiple files.

Moving to Agent mode takes this a step further – in this mode you can ask Copilot to perform more tasks autonomously.

For example, in agent mode, issuing a command like “create a script to automate building a docker image for this project” might create the Dockerfile, and then a bash or windows bat script that will execute various steps that would be required, such as executing a build tool before building the Docker image.

In agent mode, like in edit mode, you will need to review the changes made after they are generated, rather than as a preview, but this feature certainly sped up some of my actions.

4. Chat can be a two way conversation

The interaction pattern with Copilot is that you ask it to do something and it responds (or actually does it, based on the mode we just discussed). But if Copilot were another human programmer and you asked them to do something, they would likely come back with some questions for clarification. If I say “please can you create me a class to store an Address” they might ask what fields you want to include in that class for example.

The feature I discovered, is that you can tell Copilot to ask you questions in your prompts. For example, using the prompt “create a new entity for a User. Ask me questions to determine what parameters and methods should be in the class, suggesting likely candidates. Ask me one question at a time” will guide you through the creation of the required code.

In this example it asked me which fields I wanted, making some reasonable suggestions, then when I had told it, it asked me if there were going to be any relationships with other classes, and there were a couple of other questions which refined further the requirements.

This process isn’t as good as I would like it to be, but it can be helpful when you are writing code that you haven’t fully designed in your head yet – you can use Copilot to prompt you to think about what you actually want.

5. A picture is worth 100 words

The last item on this list is that you can paste an image into the copilot chat – this is particularly useful if you want to generate a user interface that looks like a pre-existing image. Copilot can analyse your image and work out how to recreate it, extract styling etc.

If you are working in an environment where a designer has provided you with mockups of what screens should look like, and you need to create an actual User Interface from these images, this should save a significant amount of effort, including thinking time.

Conclusion

By the time you read this blog post, the entries here could be out of date.  Github Copilot is evolving quickly, and there are newer features coming online at regular intervals. But the point of this blog post is to say that it can do more than you probably realised – and it can be a real productivity boost.  It really is revolutionising the way that we write code. The role of the developer is becoming one where we spend more time thinking and less time typing, but producing greater volumes of higher quality code.

Just before finishing this article I had a meeting with a colleague where they told me they had asked Copilot to produce some architecture diagrams for their project, and they were impressed with the results… so I’m off to try that next!

Success Stories

Explore our Github Copilot case studies

Success Stories

Adopting GitHub Copilot at Scale at a Tier 1 Bank

We helped a global Tier 1 bank adopt GitHub Copilot at scale to give their software engineering teams a competitive advantage.

Read success story

Success Stories

Supporting AI Learning for Technology Leaders

We built training to upskill technical leaders at an insurance technology firm, focusing on AI tools such as GitHub CoPilot, Github Advanced Security, Atlassian Intelligence, and SourceGraph Cody.

Read success story

See stories