Transcription
Imagine you've hired a new employee. They are smart, but know nothing about your company. What do you do? Of course, you give them documentation, instructions, examples. Cloud Skills are exactly the same. You upload instructions, and Claude works the way you need it to. Accurately, without errors, and according to your rules. A few weeks ago, Anthropic released a multi-page document on how to properly set up Cloud Skills, how to use them in production, how to use them in your work projects, in business projects, and more. I have translated this tool specifically for you into Russian. Therefore, if you want to get a practical tool for today's video, be sure to follow the link in the description. My Telegram channel will be waiting for you there, you will download it, but it will be your cheat sheet, so to speak, so that you can continue to work effectively with a tool like Claude. Friends, my name is Roman. You are on the Sukhov Life channel. Let's start today's episode. Let's go.
So, friends, today I don't plan to fill the air with water and talk about some theoretical part of Cloud Skills. The only thing I want to add is to emphasize a certain analogy, what Cloud Skills are. My colleague on YouTube came up with this analogy very well. I want to retell this analogy to you as well. Remember the movie The Matrix, when Neo was put in a chair and different skills were implanted into him, there were skills of karate, some languages, and so on, and so on. The same thing. we can implant these skills into our Claude tool. The same thing, let me give another analogy. For example, you are working within one chat, you have some very long prompt that you constantly drag from chat to chat. And to avoid doing this work constantly, we can implant a certain skill into our Claude. And every time we start working on some project, it already knows the context. This is the most important definition of what skills are. At first, I didn't fully understand what it was for, but then over time, as I started working more with these tools, I realized how convenient it was. Therefore, if you are not yet using skills in your Claude, be sure to try to create at least one skill after today's lesson. So let's start the practice, friends. Let's go.
So, where do we enable Cloud Skills? We expand the panel, go to the settings menu, and click the main button, Settings. Yes, these are all the main settings within our Claude. Our next task is to select the capabilities section. Click here. By the way, friends, if you opened this video by accident and haven't seen the first part yet, and you are a beginner and don't understand anything about Claude yet, be sure to watch the first episode. I talked about Claude in general, about what tools are available. We did a full 40-minute overview. We discussed the most essential tools available here in the panel in more detail. Today, we are dealing with Skills.
So, first of all, friends, what do you need to do at all before working with Claude? Within this panel, it's memory. You need to turn on all these toggles. I'll explain why this is necessary now. When you work with Claude, you regularly work on different projects, which is understandable. And it collects some summary information about you. Summary information about what it needs to remember about you. Let's even take my example, here memory from your chats. Well, in principle, it's clear what this means, memory from your chats. Let me now translate it into Russian completely, because it saves everything in memory in English. And we will see what it writes about me here. Working context Roman. Automation specialist and business owner who manages 24i, a company specializing in business process automation. In principle, you can pause and read about me. Everything is written correctly here. Even within some projects, within the last month, what I completed a large project for a law firm, it was. We worked on these projects with him. Well, and Roman has gained experience in consulting, right? So we discussed some business strategies with him, so it accumulates information here, and this information, in principle, can be corrected. So here we can, for example, write something else, add some specific information. I love cats, I don't like dogs. Don't remind me of my former baseball team. Broccoli is my favorite vegetable. Look, there are cool tips here, by the way. Therefore, if there is any important information that it should remember about you, you can write it here, add it, or just some information so that it knows about you, and it's normal that it knows about you, because it is very useful within some of your future projects when you work in other chats.
Next, what I would focus on is skill examples. Let's see what skills are already available here. Firstly, I haven't added any skills here yet. Specifically, in this chat, I have two Claude accounts. Well, I will now load the skills that I need here together with you. And let's see what's here. Algorithmic Art. Creating algorithmic art using PF5GS with seeded randomness. Brand Guide applies Antropic's official brand colors and typography. So, if you work with design, for example, this will suit you. Canvas Design. I have this thing enabled. So, create beautiful visual images in PNG documents using design philosophy. You should use this skill. So, very important, friends, MCP builder, this will be useful for you, because we will be discussing MCP in the next lesson. So enable this skill. And this is the skill creator. Today we will be working with this skill creator. This is the most important toggle to enable. So, be sure to enable it. How it will look in English, let me show you. Skill creator, yes, accordingly, you enable this toggle and, accordingly, now we will create skills. So what else do I want to add? Topic: "Factory, Web Artifact Constructor". Also an interesting thing. These are the built-in skills that you can enable if these skills resonate with your activities. So, I think there should be no questions here.
So, guys, the first skill I want to create is to transfer the work of my Telegram GPT. This is my GPT that writes posts in my style. I fed this GPT dozens of my posts, probably. And, in principle, if I start writing a topic in points, it, within the 52 model, provides some information, some post for my Telegram channel. I want to create a skill now so that Claude writes in my style. This is what I need to do. I will now transfer the GPT settings completely to my Claude. And here I have a specific instruction. You are a personal copywriter for the Sukhov Life Telegram channel. And here is written, the instruction on how it works. Yes. And in fact, I have a Telegram style guide. So now I want to download this file. It will be useful for me, because I don't have it on my computer right now. Well, this file has been downloaded. And we can now move on to the next stage. I go to Claude and, uh, skill creator. I click on the three dots here and go to Try in chat. Okay, I already have the prompt written here in English. KCL, can you make something amazing with skill creator skill? Well, version 4.6 is working for me, so I send this request. Now it's thinking, opening the skill creator skill, that is, reading the skill creator, apparently an internal document. And now it will ask me questions: "What do you want? What skill do you want?" It will write N8N workflow designer skill client protocol Whisper Skill Telegram Bot. So, within the projects that I have already used, it asks me questions. So now I will prepare the next prompt. Create a skill for writing posts for my Telegram channel Sukhov Life. I want to transfer information from my GPT, so that you study this information and format it into a skill. This is what the GPT instruction looks like. Also, an attached file with the writing style. Now I will paste the instruction and drag the file here. I send it here. So, while it's doing all this work, so that you understand, to create a skill, we just need to go to the settings section to get directly into the chat. And in fact, Claude will help you, guide you by the hand so that you can create your first skill yourself. So, if you have some idea now that you constantly transfer from chat to chat, in ChatGPT, in Claude, it doesn't matter, you can create a skill with it. Maybe your idea is already formatted as GPTs, with ready-made instructions, like mine. I would, of course, be able to upload the skill, within this Telegram Git MD, but the trick is that the formats won't match. There is a specific format for skills. And now, when it creates this skill, we will look at what this format consists of, because there are certain rules. These rules are written, by the way, in the document I mentioned. We will now analyze this structure. I will show you what the skill.md file consists of.
So, friends, it's quite interesting. It created two documents. The first document is a ZIP file ready for download with the correct structure. And you can look at the contents of Skill MD. Well, let's download all these files that it recommended to us right now. So, look, .skill is some kind of separate format for... Let's see what it looks like. Even if you look, you can see that it's the Claude logo. .skill and skill.md. So, now what do we do? We go back to our settings section and click on capabilities, yes, that's the section. And what do we need to do? Next, we need to add our skill. Click on Add. And here, as you can see, there are three options for adding your skill. This is the first, either try to do it with Claude, which we just did, or you can write your own instruction. This is already manual labor, we write it by hand. Accordingly, here is the description and instructions, but this is for those who like it, as they say. So now we will upload skill, upload our skill that Claude made for us. I drag these two files here. It's processing them now. And upload life post writer. Great. And our skill is ready. Let's translate the page into Russian, let's see what it says. Rewrites and creates posts for the Sukhov Life Telegram channel in the corporate style of Roman Sukhov. Well, let's try to test it in chat. Now, just out of curiosity, I'll try to take a piece of text about the sensational topic of OpenClow. I think you know, you follow the trends, right? And now I want to send this piece. Let's see how it transforms it into a post for the Telegram channel.
By the way, while we are generating a post here, we can also do the same thing, compare who will cope better with this task, ChatGPT or Claude, right? in terms of writing a letter with a style excerpt, so that it resembles my style as much as possible. Here you go, ChatGPT has already started its work. And yes, Macmini is being discussed like buckwheat before lockdown. Yes, by the way, excellent. This is exactly the style I like. Open Clow 213,000 stars on GitHub in a couple of weeks, to understand the scale. These are popular open-source projects that accumulate such an audience over years. Here's what has changed, it wrote. Well, and here OpenCl is when open source suddenly turns into a natural disaster. So, friends, I compared these two posts. I'll tell you this, that ChatGPT writes more broadly. Its style is such that it constantly likes to play with analogies, right? So I don't like it when it puts some expressions in quotes. This didn't go well. It's like you posted a PT project. And here, Claude has a more restrained style, and it maintains this style better and more qualitatively for me. In this case, when I don't like that ChatGPT puts some words in quotes, all this, and this style, you know, is very much ChatGPT's style, even when you give it certain writing boundaries, it still distorts the text. It distorts it in its own neural network style, which is recognizable at once. That's why I constantly write to it, for example, use fewer long dashes and don't use analogies or words in quotes. And here we have a clearer, more measured text, which I like more. Therefore, this is the kind of skill.
Let me tell you now what needs to be done to create such a skill. I have actually now fully uploaded the ready-made instructions for this skill to be created. Everything is done as simply as possible. You have several posts, for example, in your style. You feed Claude, show how you write. It already, within the framework of how to create this skill, takes some specific features, turns of phrase that you use in your text, and transforms it into a ready-made document. And then you can upload it, as I just did. Accordingly, now, when I need to write some posts, I can do it not only in ChatGPT, but also in Claude. And in fact, I will always have diverse posts in diverse styles, because models, even if you give them certain frameworks and rules, they will still do it in their own way. After 3 years of using neural networks, I can clearly distinguish where Claude writes and where ChatGPT writes. So, what can you do, as they say, everyone chooses the tool that suits them.
The next skill I want to transfer to Claude is drafting contracts based on the technical specifications that I upload, for example, within ChatGPT. I also have a ready-made document, ready-made instructions, yes, here is even an example of an act, an example of a contract, a development contract, and so on. Now we will upload all these three documents. And I will also ask Claude to create such a skill for me so that I can draft contracts, because in these contracts all my details are, for example, so that I already have a ready-made contract drafted with my clients. So I want Claude to generate documents like this for me. So what am I going to do now? I'll quickly download all these instructions. For now, I'll pause, and we'll come back later. I'll show you how I formatted it all.
Okay, great. I have returned to the same chat where we created our skills. Now I need to create a skill for writing contracts. Contracts, yes, for writing documents. Let's put it better as documents. Take all these instructions and format them into a proper skill. So, now my task is simply to upload, for example, this. You are a GPT lawyer, a contract generator. Now we will copy all this and paste it here as a separate document. Well, and directly all those three documents that were in my GPT agent, I will also transfer them to this chat. Great. I send all the instructions, and it will prepare the skill that I need to generate a ready-made document. We will then conduct testing within a commercial proposal so that it generates a document for me that I would send to the client for signing. Everything, now I have everything I need. Allow me to develop the skill correctly. Okay, now it will work on it. Let's see what we get in the end.
Okay, mega-sprint. Our skills have already been generated, so we download them. Everything, as we did before. We return to our settings, our settings and capabilities. Capabilities. Uh-huh. We add our new skills. We upload these two skills here now. Everything. See how quickly our contractor has been uploaded. This skill is ready. Accordingly, to use this skill, we go here. Look, there's a very interesting point. I'm now going from settings, to use this skill. But you don't have to. You don't have to do it from the settings themselves. You can start a conversation in chat, say that I need to generate a document. Within, well, for example, like mine. I need to generate a document. Use this skill and generate a document for me. Well, let's say, it will immediately dive into it. To test this skill, I will send one of the commercial proposals that I prepared for the client to our Claude so that it can draft a contract, a service agreement, right? So, accordingly, I send it to our Claude. And now our contract will be generated. In theory, so.
Everything, it starts reading the skill. Contract generation is now underway. I just didn't turn off the translation, as you understand. Well. Great. The skill has been read. I have already extracted most of the data from the commercial proposal. A few details are missing, gathering everything into one request. So, VPN service technical support automation project. So, I need to clarify the contract number. It asks which one. Well, let it be the nineteenth. Date of signing. Let's say today's, and the full details of the company, enter the text for now, I don't know. Leave a placeholder. Let's do that. Can aspect bur be mentioned in the portfolio? Yes, it can be mentioned. Okay. So now we will witness how it creates a real document for signing the contract. Let's wait a bit, and then we'll come back and see what happens next.
Look, let's not rush. Great. Confirm the details before generation. So, project cost, and payment terms, 50% prepayment. All correct. If yes, generate the document. Yes, all correct. Yes, all correct. So it should, let's say, generate a document for signing and an acceptance certificate. So we will have two documents, with which I work with my clients. So, I need to do this in Claude so that it generates such documents for me. Because, by the way, in ChatGPT, no matter how much they praise ChatGPT, it cannot do this. Generate quality documents. There are even connectors, but the quality still doesn't reach what I would like myself for it to work. Therefore, I have been doing all documents exclusively in Claude lately.
So, friends, well, everything is great. This is how the contract looks. City of Novosibirsk. So, the date we sign. Contractor Sukhov Roman Evgenievich. Subject of the contract, cost and payment procedure, acceptance of services, warranty and support, everything as written. So here is my contractor card, here is the client, client details. All this can be filled in manually, as they say. So, the acceptance certificate can also be downloaded. Well, it generates such documents without problems. Moreover, I no longer need to go to ChatGPT, copy, for example, what it did, then go to some Google tool, for example, or Google Drive, create new documents. All this is done automatically within Claude, so there should be no problems here.
In general, friends, regarding Cloud Skills and everything related to this whole story, when we study it in more detail, for example, code, skills in general, they will be stored on the computer in separate folders. So, when we create skills, we are now loading them into the Claude interface. But if you use Claude code, then of course, all this will be stored on the computer. You create separate folders with skills. There can be many skills. There are now entire libraries of skills. There is a lot of interesting stuff there, so in the future, of course, we will analyze this together.
Well, friends, then you will ask: "What about the built-in skills that are here?" Let's try to use design. Honestly, I have never used this skill before, for example. Let's ask it with such a prompt. Create a poster for business automation in a minimalist style. Dark background, emphasis on technology, text 24ai, automation without unnecessary words. PNG format. So, in theory, Claude should draw some logo for me now. By the way, I have never seen Claude draw any images. There are no problems with ChatGPT in this regard. And now allow me to check the availability of fonts. Here it starts to perform some specific task. I think we will get some interesting design now. What's interesting, well, I've really never seen Claude generate anything, any images, and now I want to see what comes out of it.
Well, friends, it worked for a long time on a rather simple poster. Yes, there's nothing special there, but this format of its reasoning is quite long and persistent. And here is your poster, ready. Roman Digital silence is a philosophy that I will choose for 24i. This is what it did. Well, in principle, we can correct something in this chat, say: "Make it in a different format or a different color scheme, as we did in the lesson with artifacts, for example." But there's nothing special, but in principle, as a phenomenon, as one of the skills available here, it can also be used.
Friends, well, this was a lesson on skills for Claude. I hope you liked everything and this lesson was useful for you. If it was useful, be sure to give a like to this video. And preferably, of course, a thoughtful comment. I read all comments and try to respond to all of them. If you have any questions, I will gladly answer them. Well, friends, see you in the next episode. This was Roman. Bye-bye everyone.