Transcription
In this video, I will teach you the most important things you need to know about Claude Code, whether you are a beginner, intermediate, or even advanced. Honestly, this is something I wished I could find in Arabic content when I started learning about Claude Code, because it is undoubtedly the most powerful AI tool today. But honestly, with all the power of Claude Code comes a bit of confusion on how to use it in the best possible way. That's why this video will save you all this confusion, and will give you truly important information that you can use today. So let's start. Okay, the first stage will always be downloading Claude Code. Now, the Claude company has made this very easy for us. You will find the documentation that I have here in the link below. All you need to do, if you are a Mac user, is to copy the first link you have here. You simply copy it, open your terminal, and from the terminal, just paste the link you have here, press enter, and it will download or give you the download steps very simply. But if you are a Windows user, you can download it in two ways: either on your CMD or on PowerShell. But before you download it, you must have Git for Windows downloaded. If you don't have Git, simply go to the link below, download Git for Windows 64 setup, and install it on your device. Then close it, you don't have to open it. Then go to the link you want, whether from CMD or from PowerShell. If you want to use PowerShell, go and search for something called PowerShell. Since I'm using Mac now, I don't have PowerShell, but simply, if you want to use it, search for it and paste the link. And the same thing with CMD, search for CMD and also paste the link you will get from here and paste it there, and Claude Code will download. When Claude Code downloads and finishes, if you want to open Claude Code, open it on your terminal, or on your CMD, or on your PowerShell, depending on what you are using. If you are on Windows, either CMD or PowerShell. If you are on Mac, on the terminal, simply press Claude, then type the word Claude, then it will ask you if you trust this file. Say yes, press enter, and Claude Code will open. Let's open Claude Code here and enlarge the screen so you can see. Now, as you can see in front of you, we have Claude Code open, but we don't have to just work on the terminal. Actually, the terminal is very good for advanced users, but if we want to see the files that Claude Code creates, we want to see the code it writes for us, we want to see the organization, I might want to add another IDE next to Claude Code. So I can simply use an IDE. IDE is an abbreviation, but it's to use a nice interface or a nice interface where we can put Claude Code so that we can see all the other things I told you about. What you need to do is you can use VS Code, or IntelliJ, or Cursor. They all use the same thing. For me, I prefer to use it on IntelliJ because, simply, they are all the same, plus I can have Gemini available, so I can use it when I want it later. Of course, you can simply open the terminal. If, for example, your terminal is not here, you can go from the top when you download VS Code or IntelliJ, go to Terminal, then New Terminal. This is actually the same terminal we were opening, or the same CMD we were opening. But here, you just type Claude Code, or just Claude, and it will open Claude here. You don't have to do anything extra. As I told you, there are other platforms you can use. You can even download it on the regular Claude, which is the regular Claude I showed you in the last video, and use it here. But for me, I prefer to use it on an IDE, which is, let's say, like IntelliJ or something like that. Now, when we get to the third tip, which is permissions, or the commands, or the privileges that Claude Code is allowed to use. Simply, when you see me pressing Shift + Tab, you will find that the permissions here will change from Bypass Permissions to just their default, which is Accept Edits, and we have the Plan Mode. Now, let me go through them one by one so you know. Our normal default mode, which is here, Claude Code can do everything for you. But every file it creates for you, or every file it edits for you, or anything it tries to edit on your device, it will take your permission. And this is actually a lot of wasted time. So I don't recommend using it. But at the beginning, start doing something like this with permissions, or with Accept Edits. Actually, Claude Code can do all the files for you and can edit them. But when it wants to do something related to the device, for example, building a file by itself, or deleting, for example, deleting files, or something like that, it will take your permission. When it's in this Accept Edits mode, I actually recommend starting with Accept Edits so you don't go to the next stages while you still don't understand your permissions. Then, the third thing we have here is Plan Mode. And I expect that Plan Mode is actually the strongest addition available in Claude Code. You can actually let Claude Code write a complete prompt for you instead of you having to know your prompt. You just let Claude Code write it for you. For example, if I want to tell it to build a website, let's say, I want a website for, let's say, for my work. Okay. Now, I can leave it like this, and since I have Plan Mode open, it will start asking me questions, and these questions will be about my work. So, based on that, it will then create the complete prompt. We will give this prompt to the AI, which is actually Claude Code. Simply, this is Plan Mode, and we already did a video about it in the last video. We talked about it more. So if you want to know more about it, it's in the last video. And you have the next permission, or the Bypass Permission. And this is actually the permission that only works when you have the regular Claude open, or when you want to open Claude instead of just typing Claude, which is regular Claude. You must give it Claude Bypass Permission, or Claude Skip Dangerously Skip Permission. When you give it this command, you will have the new addition, which is Bypass Permissions. When you open regular Claude, for example, if I show you regular Claude Code, Claude, when we try to do it with Shift + Tab, and go through the permissions, and see that we want Bypass, we won't find it. We must have told it from the beginning that we want to skip permissions. With this Bypass AI, it can do everything on your device, delete files, add files, edit code without asking you any questions. This is actually a powerful addition, but you need to know exactly what you are telling it. Because some people tell us that, for example, Claude Code deleted their files from their device and did stupid things on their device. But this is actually, for me, it has become a problem because I actually give it the correct prompt. If you give it a prompt with a mistake, or for example, you tell it to do things on your device, to mess with it, then you will bring disaster to it. So, honestly, I don't recommend Bypass Permission at the beginning. But I recommend that you start with Accept Edits, the things you need. And if you want Bypass Permission, for me, I am using Bypass Permission. So if you want to use something like this, then use it. This is the third tip. And the last tip we can talk about in the initial stages is the slash commands. The slash, when you put the forward slash, you will actually find that you have commands that you can give to Claude Code. And usually, you will have many commands that you can give it. So let me give you the most important commands and understand exactly what are the commands we can do with Claude Code. Simply, when you have a slash, you have commands within Claude Code itself that you can use to make Claude Code do something specific for you. For example, if I type slash model and press enter, it will tell you which model you want to use to build your work. Do you want Sonnet 4B 6? Do you want Opus? And do you want Haiku? These are the AI models. You can choose the model you want. Maybe you have other commands. Let's say, if you press escape to exit here. There are other commands like, let's say, Rewind. Rewind, actually, you can go back in time in your code. What does this mean? For example, if you gave your code, or your Claude Code, a command, and then, unfortunately, this command messed up your code and ruined it, you can go back in time with Rewind to choose, let's say, a past prompt where your Claude, or your code, was still working. We have other commands like, let's say, Context. And Context, simply, is one of the most important things in the intermediate stage that you need to know to build the best code you can. And this will lead us to the intermediate stage, which is Context Usage. Simply, the code, or Claude Code, has a context, or a space in its mind, where it can remember previous conversations. And based on that, it will give you the best code it can create for you. When it reaches a certain percentage, let's say 50% to 60% to 70%, the quality of the code it gives you, or the quality of the output it gives you, will decrease. So we always need to be aware of our context. Based on our context, we need to know when to clear the context. This is from Claude Code's mind, so it knows exactly what we want and how to do it. Always, when you look up here, you will find, let's say, here it says 8%. When I typed slash context, it gave us the complete context available. And it tells you that it has only used 8% of its mind. How do you know this? Just like I told you, slash context, once again, to see it. Just context. It will think about it, then it will give you back and tell you, here is my context. Okay, maybe you will tell me, "Oh, Abdul, I don't want to keep pressing context all the time to see my context." You need something like what you have below here, where the context is shown here. Your model is shown, what model I am using is shown, the name of your project is shown, which I call Abdul Project. Simply, you can tell your code, or Claude Code, to give you a statement line that I can put below, which will be there all the time, so I can see the context available for the AI, the name of the model I am working with, and the name of my project. Simply, I can do something like this. Or you can go to my community, where I give them the complete code, the skill that I am teaching my AI, or Claude Code, to do this, to have the context below. But simply, let me explain the context line to you, because it is one of the most important things you need to know. Context, simply, is just the way the AI thinks. What comes out here tells you that anything you can use in Claude Code will take tokens from you. Tokens, let's say, are the depth of your code, or your Claude Code. The more tokens you take, the more context usage you are using. The more context usage you use, the weaker your prompt becomes. So what you want to do is always look here. When we look at the context, we need to know that the last 80% of the context is not usable. Because Claude Code is always summarizing what it has done. So we want to stay away from this area. I told you that around 50% to 60% is the best place to use Claude Code, which is around 50% to 60%. Let's say, here, when we reach this dangerous area, which is the last 30% to 40%, you actually always want to clear your context, or clear Claude Code's mind, so you can use it again. So, simply, you can do it with slash commands, which is clear. Clear, simply, you clear Claude Code's mind and return it to zero. But now you are thinking that when we return Claude Code's mind to zero, it will forget everything we did. Actually, it will not forget everything, because it can go back and see the files you have created on the left, and based on that, it will remember what you want. So when you ask it about your code again, it will remember the code. It will not be like before, when you clear, or change stations, it will forget all your work. No, here it keeps remembering your work, and based on that, it will develop for you. So, always know when you reach around 50% to 60% of the context, always return it to zero. Because if you don't return it to zero, you will keep having problems with your prompt being very weak, and Claude Code not being able to give you a correct prompt. Now, the next point we need to talk about is Git. We talked before about Rewind. Rewind, we said what it is. Actually, it's like going back in time to the code you saved. Simply, this method is good, but it's not the original method we use. The method we use is to go and create a Git file, or a GitHub. I've told you about it many times. It's actually a place where you put your folder, or your file, and put all your code on the internet, 100% free. And from Claude Code, every time you edit a file, you tell it to save this edit for you here, then upload it to GitHub, so that the code is always saved on GitHub. You link it to other platforms later and do whatever you want. Simply, how do you do this? You have a GitHub account. You will find the link below. Go and create a free account. There is no payment or anything. Then go to your Claude, and simply tell it, when you finish your file, just tell it "Connect to my GitHub account." Then, after you do the first thing, connect, which is linking your account to GitHub, you tell it the second thing: what do you want to do? You want to save your code. So you tell it "Commit my code." Commit, simply, is just saving your code. Then "Push to the repo." Simply, the repo is an abbreviation for your folder on GitHub. So what you want to do is simply tell it "Commit" to save my code, then "Push to the repo" to upload it to my repo, or my file on GitHub. Simply, press enter, and your code will upload. I have actually uploaded the code. If you see my code that I have already created, this is actually the same code that I wrote here in IntelliJ. It's the same as what's on my GitHub. I simply tell it to upload the file here and create a file that is ready to be used anywhere, or on any host, let's say, later. Now, you have a skill that you can use in Claude Code. Actually, it's not like a point I want to give you, but it's actually a skill. When, for example, you create code and see the website, let's say, if I create a website and my localhost is not to my liking, and I want to fix its design, instead of just telling it to leave the file, or leave the code, or make the design nicer, simply, if you look here, actually, Arabic and English in the terminal are bad. If you look here, I'll tell you what you can do about this. You can simply tell it slash, then parentheses, and tell it here, "Search engine." Why do I do this? Because, simply, our AI is trained from a few months ago. Then, when it's released to the market, actually, everything new that's released after it's released to the market, it can't learn about it. So we let it use its power, which is that it can go and search for itself. Any website you can think of, based on that, it will give you the answer. It will go and search for answers, then learn from them, and based on that, it will build our code in the best way. So, instead of us telling it to create a design from nine months ago, it goes and looks for the newest things on the market, then learns them, then creates it for you on your design. Now, there's a point before we go to Protips, which are things for professionals, which is actually your mindset when using Claude Code. Now, Claude Code AI is very powerful and can do what you want. But if you just keep sitting with Claude Code and keep telling it to just press enter for anything it tells you, and accept that any file you tell it to create, or it tells you it wants to create, you accept to do it, simply, you will not benefit anything later. If you want to reach the advanced stage and be able to deal with your clients, you actually need to understand the code that is being written. I don't have to open every file for you like this. If I open this and understand exactly what this code does, or I know how to write this code exactly as it is, but at least I should understand exactly what this code is for. Why is this code for design? Why does it do this? When, for example, I look up, why is there this error? So that I know later how to solve these problems and errors that come from the code. So, simply, I honestly recommend that if you want to learn code, I'm now, let's say, I'm lost in these files and I don't know what's happening. Simply, I just go and tell Claude, I open its Plan Mode, I delete the things I told you about a little while ago, and I tell it simply, "Claude Code, I am currently lost in all the code you have given me. Can you please explain to me exactly what you did, how you did it, and what are the special things we are using in our code?" First, you do that. Then you change your mode, or change your permission from Bypass Permission to Plan Mode. Then I tell it enter. It will give me a complete plan, or the things it has actually done, and it explains them to me in detail. And this is actually the power of Claude Code. It's illogical for you to just keep accepting everything without understanding the code. At that time, when you deal with your clients, you won't know how to explain your code to them. If problems arise, you won't know anything but to go to Claude. Therefore, you should try to understand what this code is doing, even if you are a beginner. Little by little, you will learn step by step. You will reach the stage you want. Believe me, for those who noticed that when I write in Arabic on Claude Code, it's actually very bad. "Hello, peace, mercy, and blessings of God." This is present because the terminal actually doesn't accept Claude Code correctly yet, or it accepts Arabic script correctly. So I recommend that you download an extension from here, additions, which is Claude Code, regular Claude Code. When you open Claude Code, download it here. When you download it, you will have something like this. Now, you can use it like the terminal. For me, I prefer the terminal, even though the Arabic script appears badly. But if I say, for example, "Hello, peace, mercy, and blessings of God," it will give it to you, at least the Arabic script in a neat way. Exactly. So, to understand Claude Code, you can use it from here or from here. It won't affect you much, because you are already inside your code. So anything Claude Code modifies, you will be able to see your files here. But this is a point for those who use the Arabic language, so you know that we can use it like this and we can use it like that. Now, the most important thing you need to know at this stage is to know what skills are. Skills, simply, are Claude Code skills, which are the skills you can teach Claude to do a specific task in the best way possible. It might be a way that suits you personally. It might be a way that suits your work. For example, if we look at Claude Code themselves, Claude Code themselves have actually released skills or abilities that you can teach your Claude so that it can do something specific for you. For example, let's say our Claude Code needs to do front-end design, or it needs to create nice user interfaces. So I want to develop its skills. So I will give it a skill to create a nice front-end design, like the one from Claude Code themselves. Simply, if you look here, this is a skill you can teach Claude Code. If you look here in front of you, simply, if I go to Skills, Front-end Design, and go to Skills, you will find that the entire file, the skill, is actually just writing. Actually, it's a Markdown file that you can give to Claude Code so that it can do something specific in a specific way, in the best way possible. I can tell you, but if you look here, actually, the skills are just writing, as if I am taking this entire prompt and giving it to Claude Code. So when someone tells you "skills," don't expect anything too advanced. Skills are just a file that tells Claude Code what to do in the best way possible. How can you use it within Claude Code that you have here? Let's say we want to use this platform, which is front-end design. This is the skill. I want to give it to Claude Code. So I go to the terminal you are using, anywhere you might be using it, and simply type slash, and then type, for example, plugins. Plugins are just additions or skills that Claude itself has released. So we can use them and download them, and then give them to Claude Code. For example, we said we want to go to Discover, which are the platforms available. Let's say we want an addition or a skill for Supabase, so that it learns how to create organized databases. So we find Supabase. You find Supabase here. You tell it enter. It will ask you if you want me to download it. I can say, "Okay, download it." Once it's downloaded, how do you use it? Simply, you type slash, then the name of the addition you downloaded, or the name of the skill. Let's say, the front-end that I already downloaded is the front. Then, based on this code, actually, here I tell it my command or my order. Let's say, let me just escape this so I can go back here. And I tell it, for example, here, "Front-end design. I want it to fix my design." And then I tell it here, "What is the thing I want?" For example, "Fix the main screen of any page we have here." Simply, this is actually you just using your skill, plus telling it what to do with it. So it knows that it needs to use the skill. So it goes back to the skill you gave it, reviews the file, the code, or the writing you gave it, or the writing that is in its skill. Based on that, it will give you the best way it can give you. And of course, just as you can use skills, you can also create skills. Simply, you tell Claude Code what skill you want and how it works exactly. And based on that, it will create it for you. For example, let's say today you have clients, okay? And these clients, you have created an application for them. And this application goes through three stages. Let's say, the first stage is qualification, the second stage is calls, and the third stage is delivery. So you come and tell Claude here, "Okay, I want to create a skill. Let's say the skill is called 'Abdul Customers Skill,' and this skill does one, two, three." So you create it as a file and teach it to me and prepare it as a skill so I can use it later. Simply, then you let Claude Code create this skill for you. So, just as you can download skills, you can also create your own skills. And now, with this, I have given you almost the most important things you can use within Claude Code. Of course, there are other things like Claude Teams, where you can have multiple Claudes working together. And there is WorkTree, which is the new thing that has been released. It's almost the same, but in a more complicated way. And there are MCPs, which you can also use to connect other applications within Claude Code itself, so you can deal with them, like GitHub. How we connected GitHub, we can connect other applications like Supabase, Slack, all these things, but in general, this video was to give you the basic skills you need to know. So when you are using Claude Code, if you have any questions, ask me below so I can make a video about it. And if you are interested in learning more about Claude Code, I have created a masterclass within my community where I teach you exactly what I know about Claude Code. So you will find the link at the top of the description box. And that's it, peace.