Copilot CLI Tutorial #2 - Commands

Okay, then my friends. In this lesson, we're going to be looking at a few things starting with something called {slash} commands, which are a common feature across a bunch of different agentic coding tools. Our cloud code has them, Gemini CLI has them, and also Copilot has them, too. So, {slash} commands, in a nutshell, are just pre-built commands that we can run to either use a specific Copilot feature, update a setting, or change the AI model that we're using, etc. And we've already seen a couple of these in action. We've used the login command to log in to the Copilot CLI, and we've used the exit command to exit a session. So, as you might have guessed by the name, {slash} commands all start with a {slash}. And if you type a forward {slash} down here, then you're going to see a list of all the {slash} commands currently available to us, which we can key up and down through. For example, we've got this agent one for managing custom agents, which we'll talk about later in the course. And we've also got this allow all or yolo tool, which enables all permissions for the current session, and that way Copilot doesn't need to keep asking you before doing something. So, be careful about using that yolo mode. Now, I'm not going to go through every single command in this list because I think there's about 50 at the time of recording this video, and it would take me too long to look at each one, but we will look at a few now, and then as we progress through the course, we'll naturally be using others as well. So then, to begin with, I'm going to go to this change log one, which lists out all the changes made in the most recent version of the CLI. I'm going to hit tab. Now, when we do that, you're going to see it populate the chat with the command, and next to it, we can see these grayed out bits of text in square brackets. Now, these square bracket values are kind of like optional flags that we can use when we're running {slash} commands. Not all {slash} commands use them, but sometimes they do. To begin with, I'm just going to hit enter to run this command without any flags, and when I do that, we can see a list of all the changes made in the most recent version of the Copilot CLI. Okay, so now I'm going to run it again by typing change log, but this time I'm going to add on the first flag called summarize, which tells Copilot to let the AI model to provide a summary of the changes. And now, if I hit enter, we can see those changes presented to us in a slightly different way, in a summary. All right, then. So, let's just go through a few more of these commands and see what they do. So, we have got down here a copy command, which copies the latest output from the session. So, in our case, that would be all of this stuff right here, the summary of the changes. Uh if we scroll down a little bit more, we can see this delegate command. And what that does is delegate the session to the cloud. And by that, I mean it sends the session to a remote server on GitHub, where Copilot can work remotely on whatever we're working on. I'm going to talk about that later on in the course. Uh we've got the diff command right here, which allows us to review changes made in this current working directory. Uh we've got the exit command, which we've already seen. It just exits out of the current session. Uh we've got the experimental command, which shows us all the available experimental features, and we can enable those or disable them. Uh down here, we can provide feedback to the Copilot CLI. Uh we've got the help command, which is really useful, because what that does is show us all the different shortcuts. So, I don't remember all of these, and sometimes I come to the help command just to see these different shortcuts. And also, it splits up all the different commands available to us into different subheadings. So, anything to do with agents is right here, anything to do with code, permissions, the session, etc. So, this is really helpful. I'm going to exit out of the screen by pressing escape, which you can do to exit out of pretty much any screen in the CLI. And I'm going to press forward slash again. Let me go right back down here to where we were, which was on the help command. All right. So, we've got the instructions command, which allows us to view custom instructions. I'm going to show you how to do that later on. Uh we've got the logging command, which we've already seen. You can log out as well. Uh the MCP command is to do with MCP service. Again, we're going to talk about that later on. Uh the model command, I will show you in a moment. We've also got the plan command to enter into plan mode, and that is to make a plan of something kind of larger scope, I guess, that you want to implement with Copilot. Again, we will see a lot of this later on. Uh down here, we can rename sessions. Uh we can resume a current session that we already had. We can share the the current session, and what that does is create a markdown file or an HTML file, and it allows us then to share that file with someone else. Uh skills, et cetera. There's lots of different commands. Like I said, I can't go through all of these. It's really best just to have a play around with them. But, what I do want to show you is this model command, because this model command allows us to switch models. So, like I said before, we don't have one dedicated model with Copilot. We have different models from different providers. So, depending on your plan, you're going to see different models here. Now, I'm on a pro plan. And like I said before, they've paused it at the time of recording this. So, you might see less than this if you just have a free plan, but you can still choose between the different models that you see. So, by default for my sessions, I have Claude Sonnet 4.6, which I find to be a good all-rounder. It's good at most coding tasks. But, for really kind of simple tasks, I might use Claude Haiku 4.5, because that's a really lightweight model, and it generally costs less. So, I could do that if I wanted to. I could press enter on here, and you can see down here it shows the current model we're using. Now, I'm going to switch back, because like I said, most of the time I like to work with 4.6. But, I've also found, to be honest, the Codex model quite good as well. All right, so let me press enter. And by the way, Opus models used to be here as well, but because GitHub are reconfiguring all the prices, they've taken the uh the Opus models away from the Pro accounts, and they've just allowed them for the Pro Plus accounts. At the time of recording, they might come back, they might not. I'm going to keep it with 4.6 for now. Also, with certain models, we get different reasoning levels. So, you can see right here, we can choose low, medium, or high. The medium is the default, but low reasoning means that it doesn't think much about the problem, so that might be better for simple problems. Medium is the default, so it's got a balanced kind of thinking, and then high is for more complex problems, because it does a lot of thinking. But, obviously, the higher the um level of thinking, the more tokens it's going to use up. So, I'm going to stick with medium for now. All right. So, finally, in this video, I just want to mention this resource on the Copilot docs, which lists out all the different models currently available in Copilot and the different kinds of tasks that they're suited to. So, I will leave this link below the video. And if we scroll down a bit, we should see a list of all the current models available in Copilot for the different plans. And there's a little description about the kind of tasks or work that that model is best suited for. For example, you can see that it says that Chat GPT, not Chat GPT, GPT 4.1 is good for fast completions. And if we come down a bit, it says that the Codex models are good for more agentic coding workflows. And if we come down further, it says the Opus models are best suited for complex problem solving and reasoning. Now, I will say that experimenting with the different models, you'll get a better idea of what each one is best at for you. So, don't just go off this list, but it is nice to have a general idea as to the scope of the models before you start using your tokens on them. Okay. So, that's about it for this lesson. And in the next lesson, we're actually going to start using Copilot to make some code changes and add some new features to our project.