8 Claude Skills I Can’t Live Without

0:00 I believe skills is the single most important feature to learn and use in cloud right now. It's how you go from prompting a chatbot to actually automating real work. I've been building many skills over the last months, but the eight I use most aren't for specific workflows. They're what I call meta skills. Skills that upgrade clot to perform better across a wide variety of tasks. So, in this video, I'll show you the eight skills that I use almost every day, show you what problem they solve, demo each one, and how to install it for yourself. So I'll show these skills through cloud co-work but they can be used in cloud codeex or any other AI provider where you can use skills right now. Now if skills are still new to you I have a full tutorial on skills which I'll make sure to add in the description below too but they're essentially reusable instruction sets for claude on how to do a specific process or task.

0:45 We've been building uh skills to automate many of our common tasks across sales, across marketing, uh operations, and other business departments, but the ones I'll cover in this video uh aren't workflow or function specific, but make working with clot or co-work better, no matter what you do or what industry you're in. All the skills I'll show uh today, I'll add for free in the first link in the description below, so you can use them for yourself. And if you want to use or customize all our workflow specific ones that me and my team are using and building out for ourselves, plus unlimited one-on-one tech help and multiple weekly Q&A with me and my team, you can also check out my AI accelerator in the first link in the description below. Now, the first skill is one that I built myself and has made a big impact on the quality of the skills that I'm building out, which is the process interviewer skill. And what it does is simple. It basically helps

1:31 you plan the process of building a skill or any other complex task by interviewing you on the process pretty deeply with 10 to 15 plus questions. Now, the reason I really like this skill and I built this skill in the first place is because of two main reasons. First of all, in my experience, cloud and AI in general are ready to jump into action too fast without having the full context of what you're trying to do. And this is fine for simple tasks, but for more complex things like building good skills, we want these models to have a full understanding of what we're trying to do and the process we're trying to automate before taking action and building the skill. But second, which is probably a bigger point, many of our day-to-day tasks, workflows, or processes are hard for me and people in general to articulate clearly because if you think about it, many day-to-day tasks are just something we do. We've

2:18 never actually thought about the process behind it. A lot of our day-to-day tasks are sort of fuzzy or don't follow a clear process. And sometimes we also try to build a skill out of something that we actually haven't done yet or are not doing efficiently yet. And yes, of course, Cloud has the built-in skill creator skill and the planning mode. But again, in my experience, these still just jump into action far too fast, which then can often lead you to an endless loop of optimizing the skill and leads you to having dozens of skills in your account that you never actually use because they're not efficient or working well. I'm sure you can relate if you've been building uh some skills yourself. For example, I recently used this when trying to build my YouTube ideation skill, which was very helpful for me, but because I don't actually have a very clear-cut process. It's just sort of something that I come up with by looking

3:03 on YouTube, my own analytics, my past videos, but there's no clear process to it. So, I asked Claus, "Use the process interviewer skill to help me plan for building a YouTube ideation skill." And as you can see, Claude first of all identified that I actually have other YouTube ideation skills. And his first question is, what's actually broken or missing with the current setup? I then said that the current skills are not really the full process. It's not working well and it's difficult for me to articulate. It then started asking me many questions which I just answered with brain dumps through whisper flow and by just brain dumping my way through and not thinking about the process too much slowly but surely gets a better idea of what I'm trying to do in this fuzzy fuzzy process and after going through many different questions as you can see and the whole point is this might take you 30 minutes to to 45

3:49 minutes but in my experience it's fully worth it. It gives you a summary which you can then confirm and once confirmed and you have that shared agreement or understanding and it then gives you a full skill PRD or a skill brief with all the best practices on how to build good skills from entropic together with the exact process you've just defined together. You can now use this brief to start building the skill directly in the same chat or you can just copy it go into a new chat say something like I want to create a YouTube ideation skill according according to the following PRD. You paste in the PRD or the brief and cloud will then have all the context to actually build you a good skill from scratch. Now, this becomes even more powerful when you have a second brain set up because I've embedded in the skill to automatically answer questions from the second brain or the OS if the

4:35 information is already there. So, you can see it first check let me first check what's already exists in your workspace so I don't ask you questions you've already answered. You can see it actually took some context documents for my second brain like idea generation framework, YouTube engine, my YouTube strategy, etc. You can imagine the more context you have, the faster you can get to aligned and shared understanding to build good skills faster. I actually adjusted this scale based on the grill me scale from Matt PCO. So, credits to him. It's a very good scale, but I basically customize it a bit more to be more applicable to skill building and by also adding in entropics best practices for good skill building. Now the second one is the prompt master skill. This skill I found online and it's a simple one but it is a gamecher. As you can see all it does is it generates optimized

5:21 prompts according to best practices. Now why is this so good? Because if you're anything like most people in AI right now this is how your prompts look. You're just brain dumping your way through with an audio transcription tool like Whisperflow into Cloud and getting messy unstructured and errorprone prompts out to Cloud. And this is fine for simple tasks, but if we have complex tasks, actually optimizing and structuring that brain dump into a structured prompt with best practices is going to improve your outputs. And this becomes especially powerful when you let Claude apply this automatically, which I'll show you in a second. But let me show you an example. The way I recommend using it is whenever you have a complex task and you do a long brain dump is just add a line at the end that says, "Please optimize this prompt using the prompt master skill." If you have a second brain set up again, you can make this even more powerful by saying

6:06 something like, "Also, pull in context for my OS to make the prompt even more optimized." You then send that off. You then get a full optimized prompt according to best practices, which you can then use in this chat or in a new chat if you want. And because I have a second brain or an OS setup, it actually pulled some context to make the prompt even better. So, you can see it already has context around me, my voice, etc., and it embedded that into the prompt. It's one small extra step in the process, but I highly recommend doing this because your outputs will get far better. Now, one really powerful thing you can do with this too if you often use these long brain dumps and use cloth for complex task is to add this as an instruction in the general instructions in clots or in the cloud MD. For example, in co-work, you can go to the settings, go to co-work, then go to

6:52 general instructions. And here you can add in a prompt like every time I use a long unstructured prompt for a complex task, auto run the prompt master skill on the input prompt and then let cloud run with the optimize prompt. Or if you use the cloud.md or and a second brain, you can add it in the folder instructions. Every time I use a medium to long unstructured prompt for a task, auto run the prompt master skill. And you can see here I did that same one without saying that it has to use the prompt master. And it says your role about medium long unstructured prompts. I'll run this through the prompt master. So it automatically ran it through the prompt master skill before processing the task. Now the next one in my opinion is a skill everyone needs to use and is one of my most used skills which is the humanizer skill. Pretty straightforward. It says what it does. It removes signs of AI generated writing from text. And I

7:37 run this on any text that's generated by AI and goes out into the world for other people to read. Whether that's an email, a community guide, a LinkedIn post, a YouTube intro script, anything really that will be read or heard by other people, I usually run the humanizer skill on. It basically detects and fixes signs of AI writing based on Wikipedia's comprehensive signs of AI writing guide. For example, here I ran my LinkedIn writer skill to write a LinkedIn post based on my last YouTube video. I got an output and after I got the output, I asked lot now run the humanizer skill on this LinkedIn post. And you can see the changes it made to the post to make it sound more human and less AI. So replaced vague hype with something specific. Full team sync is team wide sync. Biggest unlock for me. It changed to changed everything for me. For

8:24 example, unlock is a very common AI word and goes through multiple of these changes for anything you apply this on. Here I ran it on email that I sent out to my lawyer. As you can see, it dropped the hope you're well, a very clear AI tell. And here I ran it on a new onboarding email for my AI accelerator. Again, very useful across the board. You can also slowly adapt this more and more to your writing style. If there's specific patterns or words you don't like that AI tends to use, you can add them into this skill and customize it to your own version. And another tip I highly recommend which is if you're using copywriting skills. For example, I have a LinkedIn writer skill and a newsletter writer skill. Add a step in your old skill to always use and run the humanizer skill on the last step before giving you the output. And the way you

9:10 can do that is by just telling Claude something like can you update my LinkedIn post skill to always run the humanizer skill after it's generated the initial output. It then updates your skill to add a step in your skill to use the humanizer skill before giving you the output. But again, anything that's AI generated that goes out into the world, in my opinion, you have to run the humanizer skill on top of it first. The next one actually builds on this. Besides making sure AI outputs sound more human, the fact checker skill actually checks if what is said is factually correct. It does a systematic fact verification and misinformation identification using evidence-based analysis. Now, the three main use cases that I use this for is first to check if what AI is generated is actually factually correct. Again, highly recommend using it on everything that's

9:55 AI generated and you put out into the world. For example, I can run this on a LinkedIn post after it's generated. You can see it first reads the current post and systematically verifies every factual claim in it. You can see it reads the current post and systematically verifies every factual claim in it. In this case, it's based on my video transcript. So, it actually checks if it's also according to what I actually said in my video transcript. And second, it cross references and looks for external verification sources by searching the web. It then gives you a full fact check report with a summary. In this case, five are completely true, two are mostly true, one is unverifiable, and zero are completely false. Overall, this post is factually clean. Now again, you can of course add this skill as a step into your copywriting skills to always use the

10:42 fact check before generating the final output. Now the second use case I use this for is to check if what I came up with and put out in the world is actually factually correct. For example, I often use this on my video outline. I write down bullet points of things I'm going to cover in my video and I just want to make sure and fact check that the points I'm mentioning is actually factually correct. So here I said, "Can you check the new YouTube video outline in notion on cloud routines and use the fact checker skill to make sure that what I'm covering is actually true." He then went through my notion, read the full outline and gave me two meaningful errors to fix. For example, I mentioned something about GitHub trigger events and it said it was false because it understates the capability by a lot and it even suggests a fix right away. Second, I mentioned that a limitation is the 1hour session timeout paron, but

11:29 there isn't any official entropic doc that actually mentions this. And for this, it's been really useful because sometimes I make points in my videos and I just want to make sure that what I say is actually correct. And it does take in the context of my video also to give me recommendations on maybe how to phrase things better or in a more nuanced way. And of course, the third use case is to actually fact check someone else's post or article or anything else and quickly check if what's stated is actually correct. So here I threw in an article. We can also add, of course, one specific point that you see somewhere, etc. Again, I think a key skill to have in your arsenal, especially with more and more AI slop being generated and being out there in the world. Now, the next one is another real meta skill, which is the find skills skill. Again, great skill to have, and it does what it says it does. It basically helps you discover

12:15 other people's skills from the skills.sh ecosystem. Uh they have more than 91,000 skills uh with the amount of installs. Now definitely worth exploring when you're trying to do a task that you don't you might not know how to do yet or how to do well yet because other people likely have built uh some skill around it. For example, here I said use the find skills skill that helps me with writing good LinkedIn ads. And they found several skills that help with that and with the amount of installs. So you can uh find and choose the most popular one and then cloud can directly install it for you if you just ask it to or you can use the command line here which you can paste in the terminal if you use cloud code. So again great skill to have and definitely worth using for any task that you don't know how to do well yet. Now the next one is another big one which I use almost every single day

13:00 which is the front-end slide skill. Now if you have played around with cloud code or cloud co-work you know that can now generate um an HTML generated presentations pretty quickly but the quality of the output is not always good enough to really use as a presentation in your company or in a YouTube video or in a sales pitch. Um and the front end slides scale really fixes that immediately. You can basically build animated HTML presentations from scratch or by converting PowerPoint or PPT files into animated HTML presentations. Now this skill already has all the best practices for good slide layouts embedded. So the quality of your slides will improve instantly when you use this skill. Now here I just use an example. Use the front-end slide scale to create an explainer deck on the evolution of AI since the beginning to April 2026. And

13:47 you can see it reads uh skill ND and gets the style presets. Important to note is if you don't throw in your own uh brand style or brand guide and you do something around AI, it's going to choose one that best fits. So in this case with AI it will almost always choose neon cyber which is just a fixed color scheme and font for that style. But this becomes far more powerful when you throw in your own brand guideline which I did here. Can you run this skill again? Keep in mind the visual identity laid out in the attached document. So now it applied my brand colors, typography and design roles and then generated an animated presentation slide deck according to my style and brand. And again, this is really powerful and you can use now for anything from personalized sales proposals inside of sales calls or any asset you want to

14:32 share with the world because you can instantly spin up this personalized animated almost landing page or presentation. For example, I'm going to implement and incorporate this more and more into my own YouTube videos to generate the visuals for my YouTube videos a lot faster and in my brand. And again, really important to customize this one to your brand. So you can just tell cloth customize the font slide scale, throw in your color scheme, your fonts, etc. your brand guideline and it will automatically update it and make it personalized for you. Now one more which is worth checking is the decision toolkit skill. I've been personally using cloth uh especially with my second brain set up more and more for strategy use cases and as a strategic sparring partner because with the context Claude now has through my second brain around me and my business, it really can start giving me good advice in my experience.

15:18 Now, if you don't know how to set up this second brain yet, I recently did a full tutorial on it, which I'll also make sure to link in the description below. But this decision toolkit can help you make big decisions better by going through a systematic analysis and first uh principle thinking framework. So, it doesn't actually decide things for you. As you can see here, the principle is about guiding don't decide. For example, here I ran it on a decision if I want to increase the price on my AI accelerator or my AI community. And then it generated a decision framework in markdown but also an interactive wizard in HTML. And basically this just helps you go through the decision and the questions that you actually need to answer to make a decision. So for example frame the decision he actually identified that there are two decisions here. Should I raise prices right now

16:04 and how should I roll it out? In the second one the start fresh test. So it's a framework how you could think about it. If you launch this community today from scratch, would you pick the same price? So, it's an interesting question to ask yourself when you're making this decision. It goes through timing, who is affected, it goes through bias checks, so you can think through is this actually affecting my decision- making here, opportunity costs, scenarios, and probabilities. It can then also give you a verdict if you actually select these. So, can be quite an interesting way when you have a big decision to sort of wrap your head around it and get a better idea of making a decision. You can also get it in a markdown file as you can see here. And for example, here we also have a first principle test. What problems does raising prices solve? Can I solve the same problem without raising prices?

16:49 So basically can be helpful to just wrap your head around making a big decision maybe a little bit easier. And then lastly, this is a big one that you really need if you have trouble connecting your software to co-work or cloud code that doesn't have an MCP or a built-in connector yet, and that is the MCP builder skill. And this is actually a built-in skill from Entropic, but many people don't know about it. And it's amazing because you can instantly build your own MCP server to make cloth connect to your software. Now, if you don't know MCP, it's basically a protocol that makes it really easy for AI agents on cloth code and co-work to connect to your software to take actions and get data from them. For example, my community runs on Circle and Circle doesn't have an MCP yet. So, just build my own using the MCP builder skill. And now my AI agent can post in my community, insert threads, uh post new

17:36 documents, analyze common problems or issues or whatever I need to do in the software. And all I did was ask lot to use the MCP builder skill to build me an MCP server for Circle.so. It then asked me some questions about the capabilities and it went through the API documentation of Circle and it gave me the full MCP server together with a full step-by-step on how to install it. Generally what you need to do is you'll get this JSON file which you need to add in the cloud config file. Now where do you find that? You can go to your settings go to developer and here you can click edit config. That's a JSON file that you can open with any text file or the file editor and the JSON that it generated you need to add in here. So if you don't know how to work with JSON you can just throw this in to claude ask it to add the new JSON file

18:22 to it and once you do this the connector will be added to your connector list. I've also used this to build an MCP out of Reddit, uh, which helps me with YouTube iation. But you can use this for any software that does have APIs but doesn't have an MCP yet, which I can tell you are many softwares still. Now, that's it for this video. I hope these were helpful. I highly recommend exploring some of these because they really made a big impact on how I work with AI. You can just download the full plugin with our meta skills and import it into your own account in the link in the description below. I've also added a few more skills that are useful, but a little bit more specific. There's another agent browser skill that's generally more efficient than cloth uh browser use. They have an audio transcriber that can try trans transcribe any video or audio on any website in any media player. There's also an open router scale where you can

19:07 actually test outputs based on different N&M providers like Gemini and OpenAI and some other ones. And again, if you want access to all the skills and plugins we are building out internally for specific workflows and functions across marketing sales operations etc. You can check out my AI accelerator in the first link in the description below where we list all of them for you to use and customize for yourself. We also offer unlimited one-on-one live tech help. You can book in a call whenever you want to help you fix any issues, plan a new solution. We also have multiple weekly Q&As with me and my team uh to help you on strategy or anything else you need. So, if that's interesting to you, definitely check it out. And if you want to learn how to actually optimize skills and customize them to your specific need, which is a very important part of using skills, you can check out the video here above.