Everyday AI Podcast – An AI and ChatGPT Podcast - Claude Skills: How to build Custom Agentic Abilities for beginners

Episode Date: December 10, 2025

Capabilities? Through the roof? Usage? Ground floor.Claude Agent Skills might be one of the most useful features of any front-end LLM. Yet....it's crickets in terms of chat around it. For this 'AI at Work on Wednesday' episode, we're breaking it down for beginners and will have you spinning up your own Claude Agent Skills in no time. Claude Skills: How to build Custom Agentic Abilities for beginners -- An Everyday AI Chat with Jordan WilsonNewsletter: Sign up for our free daily newsletterMore on this Episode: Episode PageJoin the discussion:Thoughts on this? Join the convo and connect with other AI leaders on LinkedIn.Upcoming Episodes: Check out the upcoming Everyday AI Livestream lineupWebsite: YourEverydayAI.comEmail The Show: [e-mail] with Jordan on LinkedInTopics Covered in This Episode:Claude Skills Agentic Features OverviewDifferences: Claude Skills vs. GPTs vs. GEMSModular Agentic Workflow File StructureStep-by-Step Guide: Building Claude SkillsClaude Skills YAML/Markdown Setup ProcessTesting and Validating Custom Claude SkillsAdvanced Capabilities: Executable Code & Sub-AgentsCommon Troubleshooting for Claude Skills CreationTimestamps:00:00 "Claude Skill Library Unveiled"06:27 "Claude Skills Explained"07:29 Custom GPTs and Gems Explained11:18 Claude Skills vs Projects17:31 "Refining Skill Triggers Effectively"20:17 "Beginner Cloud Skills Best Practices"23:39 "Preferring GPT and Memory Tools"25:54 "Saving Skill File Properly"28:09 Creating Skills on Claude33:43 "Creating AI News Searcher"35:36 Claude Skills Now Available37:39 "Optimizing Claude for Knowledge Tasks"41:05 "Skill Builder Library Access"Keywords:Claude skills, Claude agent skills, custom agentic abilities, large language model, agentic workflows, specialized tasks, coding capabilities, file creation, executable code, skills library, skill builder, skill creator, markdown file, skill.md, folder structure, YAML front matter, composable skills, modular instructions, automation, prompt engineering, skill triggers, skill testing, advanced features, API skill versioning, governance and efficiency,Send Everyday AI and Jordan a text message. (We can't reply back unless you leave contact info)

This is the Everyday AI Show, the everyday podcast where we simplify AI and bring its power to your fingertips. Listen daily for practical advice to boost your career, business, and everyday life. Meet Firefly AI Assistant, now live in Adobe Firefly, the all-in-one creative AI studio. Just describe what you want to create and the assistant handles the rest, orchestrating multi-step workflows across Photoshop, Premiere Express, and more in one conversational interface. You direct the outcome, the assistant accelerates execution. This might be one of the most robustly flexible and capable agentic features on any front-end large language model that so few people have actually used. I'm talking about clawed agent skills.

Claude agent, what? Yeah, I see so few people talking about clawed skills, which is pretty shocking considering how, how good Claude's models are, especially when it comes to coding and agentic behavior. And all you got to do is write a little bit of markdown. And I think that's exactly where people get lost, those scary buzzwords, agentic models, coding, markdown. Well, don't worry. We're to make it dead simple on today's episode of Everyday AI and give you a walkthrough

on Claude skills and do a very simple for beginners guide on how, they work, how to build them, and how I think you should use them. All right, I'm excited for today's episode. Hope you are too. Welcome to Everyday AI. What's going on, y'all? My name's Jordan Wilson, and well, this thing's for you. It is your daily live stream podcast and free daily newsletter, helping everyday business leaders like you and me make sense of those scary AI buzzwords. We demystify it for the average, everyday, non-technical person. So if that's what you're trying to do, learn and actually leverage AI, you've come to the right place. It starts here with the unedited, unscripted live stream podcast.

But to take it on the next level, make sure you go sign up for our free daily newsletter at your everyday AI.com. We're going to be recapping today's episode. So if you missed something specific, maybe you're walking your dog or running on a treadmill or running from a dog on a treadmill, I don't know. But we're going to be recapping all of the important takeaways in today's newsletter. All right. And if you want the AI news, yeah, we cover that every day in the newsletter as well.

Also, Claude skills extremely impressive. But I think one of the downsets, and I think I run into this all the time with different powerful features and large language models, is it's almost like there's too many capabilities. Like, where do I start? Or maybe how do I do this, right? Even though this is a walkthrough and tutorial for beginners, you might need just a little extra help. So that's why, well, on, I'm going to tell you this. right now. At the end of the show, I'm going to tell you, we put together a skill library. This is what I do every day. I talk to, you know, business leaders, the people who make these

models, and I spend all day finding and testing different use cases. So we put together a skill library, so a Claude Skill library with 50 of the most, I think, flexible and high leverage skills, kind of skill ideas. And then we also put together a GPT builder, sorry, a skill builder inside of chat dbt we have a gemini version as well so even if you don't know how to turn a skill you know into the right format using markdown we have it there for you you can just type out your idea and it's going to spit it out exactly we've taught this thing and trained it on best practices so stick around and we're going to tell you how to get that at the end but on today's show we're going to define agent skills and explain how they turn clod into a specialized expert

we're to break down the file structure and compare skills to projects, prompts, and MCPs, well, at least GPDs and gems as well. And then we're going to walk through the step-by-step workflow live. What could go wrong doing this live? And I'm going to show you and teach you how to both build, test, and deploy your first skill inside of Claude. All right, so FYI, you do have to have a paid Claude plan to do this. So just getting that out of the way right now. But maybe this is even worth it to, if you're on the free plan, just to do this once or twice, just to even see if the paid plan might be worth it for you. All right.

So here's an overview. Well, skills, I'm going to break down kind of the differences between GPTs and gems because I think those are what maybe most people are familiar with. But skills are essentially modular folders of instructions that teach Claude specialized tasks. So they can act as almost custom onboarding materials to make Claude a specialist for you. And this feature unlocks agenic workflows by allowing cloud to execute real code, right? I think when you think of using front end large volumes models, and if you're brand new here, what do I mean when I say that, right?

So that means you're going into clod.aI and you're chatting with the Claude chatbot, right? You're not using the backend Claude Anthropic API. You're not using cloud code, anything like that. We are using this on the front end, on a chat bot, right? And Claude is extremely impressive on the agentic side, right? It's new Claude Opus 4.5, at least when you look at benchmarks. It does best in software development. It does best in coding and in agentic flows, right?

So it does very, very good in creating skills. But you have to teach them the skills. And you have to do it, you know, format it in a certain way. So a little more technical and a little more hands-on than if you're used to using, you know, GPTs or gems, which are customized versions of ChadGBT and Google Gemini respectively. But I think technically the capabilities are a little bit more because Claude has things like file creation, right? It's able to create and render code. So you can do so many things, especially when you stack these, these different skills.

So here is how Anthropic describes agent skills when they introduced it from their website. So they said Claude can now use skills to improve how it performs specific tasks. Skills are folders that include instructions, scripts, and resources that Claude can load when needed. Claude will only access a skill when it's relevant to the task at hand. And when used, skills make Claude better at specialized tasks like working with Excel or following your organization. brand guidelines. So yeah, they're kind of like GPs, but way more advanced and just way more capabilities and just more steerable overall. So let's do a little bit closer of a comparison here on Claude Skills versus GPs or GEMS. So Claude skills a little different. They're

automatically invoked by Claude for whatever relevant task that may be coming up in the course of the chat and they are focused more on governance and efficiency. So what does that mean? Automatically invote. You aren't, you know, like a gem inside Google Gemini. You're going to go in to that gem, right? You manually create it in the interface of Google Gemini or in Chad TBT, you go into the chat GPT interface and you create a GPT, right?

A customized version of Chad GPT, right? And then when you want to use that GPT or that gem, you navigate to the section inside the user interface and you say, oh, here's my little custom version of, you know, chat GBT that I did to, you know, rewrite something. Or here's my gem that I use to, you know, build, you know, visualize something, right? So GBTs and gems also use the built-in features of their,

you know, kind of parent large language model. skills different. They're more reusable folders of expertise that are loaded as needed across conversations. So you're not calling them, right? You're not going into a certain area or certain user interface within Cloud to use the skills. You have to enable them. You have to load them in. I'm going to show you how to do that.

And then Claude just kind of decides agentically, right? You can call it out my name, which usually helps, right? If you're struggling to get it to use, a certain skill, you can still call it out by name, which usually works, but not always. Right. And so a couple other differences here. GPDs and gems require manual user selection, and they're more built for specialized tasks. And then GBTs and gems also offer flexibility, broad use cases and integrated tools like

web search, image gen, co-generator, etc. So all the features, right, of ChadGPT or Google Gemini, they are available. when you are using a GPT or a gem. So a little different. The way I like to think of it is skills are kind of built outside of Claude and they are not directly used by manually selecting them. There's a lot of crossover, right? But I'm trying to explain how they're different.

But there's similarities in a lot of ways, right? because I think a lot of times, you know, as an example, Chad GPT has projects. Google Gemini is rolling out projects and Claude has had projects. So people thought like, oh, Claude, well, that's kind of like a GPT or a gem. And it's like, no, they didn't really have anything until they rolled out skills, but a little more technical and a little more setup. So let's go over kind of the technical side of how they work.

So Claude essentially scans your available skills, anything that you load in. And infraffic does have some built-in skills that you can either enable or disable by default. So it goes through and it matches the relevant tools to anything that you're talking about in the Claude conversation. Then, excuse me, still this sore throat, get me, not leaving me alone. So then the metadata loads first to keep Claude fast while accessing specialized expertise and skills kind of composable. You know, right? They're modular. They stack together. So, you know, that's nice. Because if you have maybe dozens of skills and maybe they can be used in conjunction or in succession, right? That's another huge benefit to Claude skills, right? I look back at like GPT or sorry, plugins out of chat GPD. And one thing

I loved about plugins is you could use three of them at any time. And then if you kind of prompt it the right way, Chad GPT could go and use those three plugins at the same time and pass information, you know, from plugin A to plug-in B. So right now, you can't do that automatically with GPs or gems, but you can't do that in a way with these Claude skills because they are stackable. So where do skills fit in the Claude ecosystem? So technically, you could use clawed projects to do some of these things, right, to automatically trigger any built-in Claude feature. So projects, well, if you haven't used them, think of them like an organizational structure.

So you can put all kinds of chats inside of a Claude project. But in there, you can also have files and custom instructions. So in that custom instruction, you can essentially do some of the things that you could do in Claude Skills. However, skills teach Claude how to do things. So it is a little bit different, you know, a system prompt versus teaching a model a skill that it can repeat and use when it needs to. So prompts or, you know, special instructions inside Claude projects are more for one-time requests in skills are for repeatable and persistent procedural knowledge. And then there's MCPs as well.

So MCPs connects Claude to data sources, third parties, right? They're kind of like APIs. That's not what skills are. Skills, you are essentially loading in some functionality in a markdown file and teaching Claude how to do those tasks, right? So let's talk a little bit more about some of the setup. How is a skill setup? How does it work?

What is its anatomy? Well, it is simple. Think of it like a computer file, right? You have a computer that has a hard drive and then there's folders and then there's files. Similarly, it's kind of how skills work. Skills have directories and in that directory there has to be a skill.md file. And the first, the file uses YAML or yet another markdown language front matter to hold the required name and description fields. I'm going to show you what that means here in a little bit.

And then folders can also house reference files, executable scripts, and resource templates. And the computer analogy is actually important here because you are going to be building skills on your actual computer, right? So this is not something you're just going to build in the interface. So there are some technical things, all right? And I'm going to go through them a little bit more in depth here. So you need to make sure that you name everything correctly, because if not, you're going to have some problems with your skill functioning. So you need to write descriptions from Claude's perspective, focusing on specific capabilities and clear triggers and just good prompt engineering 101 anyways, right? Avoid vague phrasing and use specific verbs like extract, create, or merge.

Here is your step-by-step workflow. Like I said, we're going to talk about this, and then I'm going to show you and do it live. So you need to start by clarifying the specific problem in defining what success looks like. So this is when you are creating your skills file. Then you can use the skill creator skill to interactively generate your folder, structure, and files. It actually should have clarified there's two different ways that you can go about building these skills. So you can actually just do it native. in the chat with Claude, which I know some people prefer.

For me, not so much. I know this might sound weird. I like making skills in other large language models. And I'm going to show you one of the reasons why and why I actually went about and created a custom GPT. I think it might make sense, especially if you're not a power cloud user. And maybe you use chat GPT or Gemini a little bit more. But essentially, you can chat with Claude and create the skill within Claude.

So you don't even have to, you know, use a third party. And when Claude did, or sorry, when Anthropic did release this skill, there wasn't a way to just save it. You still had to download it as a zip file, upload it. Now luckily, they've kind of made it a little easier. So you can, number one, you can write these skills by hand. You can, you know, use a different large language model. That's what I like to do. then do a download upload process or you can just talk natively with Claude,

kind of what they call a skill builder, right, which is essentially there's no certain place in Claude, there's no button you click, you just say, hey, I want a skill that does this, you know, use the skill creator and, you know, build this for me, right? It's actually funny. I'm pretty sure the skill creator in Claude is actually ages. default skill that they've built that helps you create skills. Kind of meta, right? Okay.

So, and here's how you need to validate them because this is where does get a little tricky and where there's kind of downsides compared to GPTs or gems, which you proactively choose and manually go in, you know, enter into the settings. It's cool that Claude can just know, oh, maybe I should trigger a skill. but sometimes you might not want it to. Right? And I've run into this actually a lot more than I thought I would. So what you should do is create a test matrix covering your normal operations of how you might

want to use a skill, edge cases, and also out of scope requests. Then you just test it, right? Start using Claude, right? So think of areas where, hey, it should definitely trigger a skill when I say something like this. It shouldn't trigger a skill when I'm doing C, you know, it should when I'm doing A, B, but not when I'm doing C, D. and test those things repeatedly and see if it triggers the skill or not.

And another thing, if you have a lot of related skills, or let's just say you use Claude all the time for one certain type of work and maybe you have 10 different skills related to that one type of project, you might get a lot of misfiring. So you do have to in building your skills, whether in the markdown file or if you're using the skill creator inside Claude, you need to be very specific. That's why I said a couple of minutes a minutes ago, you have to use kind of prompt engineering basics. You have to be extremely clear and specific using vague language. You know, skills actually might be more frustrating than they are helpful.

I think for me, you know, when these first dropped and I was using them the first couple of times, I was getting frustrated because there were sometimes I wanted a skill to enable and it wasn't. And then sometimes, you know, vice versa, it was happening. Or, you know, I do have a lot of similar skills in some of my accounts that are related. So, you know, sometimes, you know, I wanted, you know, I don't know, AI News Researcher A, and it was kicking off AI News Researcher B. So you do have to really go through and test these tasks that seem related, but shouldn't trigger your skill to verify the boundaries and then refine your descriptions.

if triggering is inconsistent or instructions if your outputs very unexpectedly. So, yeah, you might have to go revisit that skills.md, that markdown file and actually update it. So, yeah, if this makes sense, right? So you are creating a markdown file with all of your instructions, but you're not going to see that, right? The difference is in GBT's or gems, you're going to see that, right? And you can go in and easily change it. Not so, not as easy, and it's not quite the same thing working within Claude with these skills. All right.

So let's talk about some of the advanced capabilities before we jump into a live demo. So skills can include executable code for tasks where programming is more reliable. And I think that's one of the huge benefits. Also, subagents in Claude code can utilize skills to handle discrete specialized tasks. And developers can use the API to programmatically control custom skill versioning. So yeah, these are maybe things more on the advanced end. Obviously, skills can be used not just on the front end. They can be used on the back end as well in Claude code on the API side.

So some of these are more on the advanced side. But in our demo today, it's for beginners. So we're just going to be working with the front end of Claude. So a couple best practices. Stick to trusted sources, right? So a lot of people are sharing Claude skills online. I probably wouldn't use those unless you know exactly what they do. The same thing with using.

you know, third-party GPs from the GPT store. If you don't know them and if you, you know, don't trust them, you might not want to use them and hand over your data. You should also never hardcode any sensitive information like API keys or passwords directly into a skill.md file. Also, keep your skills focused on single workflows to ensure they can pose better together. Again, remember, they're modular, right? little pieces that can work and fit together. So I think a lot of mistakes that people make when

trying to do skills, same thing with GPs or gems or anything, is they just try to do one thing to automate a large project. No, they're modular. Break it down into little chunks. You're going to have a much higher degree of success if you make them small and specific. All right. So before we go live. Let's first take a very quick break for me to take a drink and a word from our sponsors. This podcast is sponsored by Google. Hey folks, I'm Amar, product and design lead at Google DeepMind. Have you ever wanted to build an app for yourself, your friends, or finally launch that side project you've been dreaming about? Now you can bring any idea to life, no coding background required with Gemini 3 in Google AI Studio. It's called vibe coding and we're making it dead simple.

Just describe your app and Gemini will wire up the right model. for you so you can focus on your creative vision. Head to AI.studio slash build to create your first app. All right. So let's get into it and let's look live. This is putting AI to work on Wednesdays, our Wednesday segment. We go live. We get hands on.

So let's do it. This one is going to be a little bit different because like I said, I'm going to show you hopefully here, you know, running live generative AI demos sometimes doesn't go very well. So if you are listening on the podcast, we always have the video version on our website at your EverydayAI.com. Hopefully it all works. Hopefully I don't bust through my clawed limit in two prompts, but we'll see.

But I'm going to be sharing my entire desktop here because like I said, I'm going to show you a very simple way on how you can build these skills across different large language models, which I'm going to tell you why you need to do it here, but also how to save them on your desktop, how to upload them, all that good stuff. So now, live stream audience, if you could, let me know. You should see my whole desktop here. All right. Not a lot going on, but I have a Chrome tab here.

I have a text edit. So I'm on a Mac. So it's just a text file editor. And then I have my, my desktop where I'm going to be saving some files. So here is why I love using other large language models to create Claude skills. they know more. They know more about me. I don't use Claude a lot, right?

Obviously, I use it every day. Right. When I say I don't use Claude a lot, I use Claude probably, I don't know, 30 minutes to an hour every single day, right? But I use Chad Givt and Gemini for multiple hours a day and I've used them for much longer. So memory, right? Chad GPD has great memory. So I can just go into chat GPD and be like, hey, based on everything you know about me,

what are 20 skills, Claude skills that I should be using, right? And it can go out and see what Claude skills are and go look at all my different conversations and reference my past chat history. And then I can use a very powerful model like GBT51 thinking or GPT51 Pro

to write some really good instructions. So that's what I did here. So on my screen, I have some instructions, right? I had it right. it, I had to write it for me in Markdown. So if you don't know what that is, it doesn't matter. It's just a simple programming, text editing language. That's all it is. So if you ask Chat Chitipati to give it to you Markdown, pretty simple. All you got to do is click copy code,

and that's what I've done. So big advantage. You can see, yeah, you can use the Claude Skill Creator, but I like using for me anyways. I like using ChatGPT or Google because that's where I do the overwhelming majority of my work. All right. So now I'm going into text edit. Okay. So all I'm going to do is paste that in. Next, uh, this one, uh, more of a little, uh, Mac, uh, hack here.

All right. And these instructions, they're simple, right? They're plain, plain English, right? So you can write these yourself, uh, but like I said, I like using other large language models and I'm doing a super simple one just to demo. This is a podcast show notes formatter. There's a description in there. There's a purpose, how to use the skill, behavior, example, et cetera.

I'm not going to read through it. It's simple, right? It's like a custom instruction that you would give a project or a GPT or a gem. It's just in Markdown. All right. So now I'm going to remove the formatting. All right. I just do this as an extra step because sometimes, especially when working with

markdown, if the formatting gets changed, something might not work. So all I'm doing is I just removed the formatting. There's just a you can go into the menu and do that. I just do a keyboard shortcut. Now I'm going to save this file. So here is where you just have to make sure to do this part correctly. I need to save this as a skill dot MD file, right? So depending on what text editor you're using, you just have to make sure to override the default settings, right?

Because in text edit, it's going to save it as a dot txt, which will not work. So I'm going to go to save. All right, I'm going to save this on my desktop. And I'm just going to, actually, I'm going to put it in a folder, right? Because this folder, again, we need a structure here. So I'm going to create a new folder. And I'm just going to call this podcast show note skill. Okay.

That part's important. And then in that folder, I am saving this file. All right. And all it is is skill.md. All right. So you are putting your own extension on there. Skill.md. All right. There we go.

So now I have my skill. MD file. I'm going to go ahead and close this. So now kind of small here, but I have this folder on my desktop. It is podcast show note skill. When I click it, it's just a skill. md file.

Very simple here. All I'm going to do, excuse me, I'm going to right-click this and click compress. All that does is it creates a zip file, and this is how I upload it into Claude. So again, let me just quickly recap what we did. We went into ChatGBT. I gave it some information before on, you know, what Claude skills are, best practices to write them. I use ChatGPT's memory to say, hey, suggest me some skills that I could use from Claude, had to write it for me in Markdown. I copied it from chat chbtee.

I pasted it into a text edit file. I removed the formatting. Then I save that as a skill.md in a folder that is a description. That description is important because that description of the folder is ultimately what the skill will be called. Then I just right clicked it, compressed it. That's how you do it on the Mac to create a dot zip file. Now from here, pretty straightforward. I'm going to go into Claude and go into Scycule.

settings. All right. And then once you go into settings, you need to go to capabilities. And like I said, you do need to be on a paid plan for this. Scroll down and you need to make sure that you're under these skills section. So you'll see that it does have some default skills. Yeah, like this skill creator is a skill. Right. So you can toggle those on or off by default, the ones that Anthropic includes if you want them or not. So you have this section here that says upload skill. Now I'm going to do, it gives you a little place to click. And then I'm going to upload. Let me find it there.

There it is. I'm going to upload my podcast show note skill. Zip. I'm going to click open. Let's see. Okay, here we go. I already got an error that I forgot. So it says skill name in skill.md can only contain lowercase letters, numbers, and hyphens.

All right. So it looks like I need to change. the name there. I swear I've done this before without the without this issue. So I think I have to add dashes between I think it was the the spaces. So I'm just going to go ahead and redo this here. Live. Love doing this live. All right. So I'm going to recompress it. Let's do it here. There we go. I'm going to reupload this skill. Oh, fun times, fun times. All right. One important caveat. here struggling with, I have to make sure that there's no capital letters in the skill name. So I'm going to go ahead and open this same file here.

It's suggested that I just use. Okay, so I'm going to make sure I don't have any capital letters and it suggested using hyphens instead. All right. So that's weird. I've uploaded ones with. capital letters before in the name. Not an issue. Love, love, love, love.

When live Gen A.I demos do this. So now all I did is I went into that folder. I made sure that the skill is not capitalized. And I just used hyphens in the skill name inside the file. All right. So now hopefully this should work well. So I'm recompens. pressing this. All right. Now I'm going to go ahead. I'm going to refresh my my

Claude over here. I'm going to upload the skill. Give it a second. There we go. All right. And it worked this time. Yay. All right. So now to use it, what do you do? Well, you don't have to do anything. But I am going to still for the sake of a demo. I'm just going to go into Claude. And I'm going to say using, I'm going to say use the podcast show notes formatter skill on this. All right. And all I'm going to do is I'm going to go to the website because this skill is made

to format podcast transcripts in a different, in a certain way, right? You can obviously make any skill. And I would suggest, you know, obviously make ones that are a little bit more useful because any large language model could do this. but I'm just doing this as a demo of this skill. Right. So you'll see I didn't have to click anything. It just knows on its own.

And I'll kind of walk our podcast-only audience through this. So it says thought process. The user wants to use the podcast show notes formatter skill on the provided podcast transcript. Let me first check what the skill does by viewing it. So you'll see kind of this agentic nature. And why, especially if you have multiple skills, right? Let's say you take a repetitive task that you do over and over, right? Going back to my plugin analogy from earlier, you could have three different plugins.

Well, you could have five different skills, uh, you know, all related to one project. And it can pass information on to each one. It can run code in, you know, step one through five. It can, um, create files, right? That's another thing that Claude is actually great at. In this case, it went through, did a really good job, um, and actually just did the skill that I trained it to in the file, right? So if I open up my, my file that I made, my skill markdown file, right? All it's supposed to do, I told her what to do. It needs to take the

raw show notes. It needs to give me bullet points. It needs to give me an episode title, short summary, key segments with bullets, right? And it's all right there. So the transcript, I paced it in there. It went through, followed my skill that I traded on to a T gave me exactly what I needed. All right. Great. So let me quickly show another way that we can do this. Let's go ahead. I started this before the show started because it was going to take a while and I didn't want it to have my screen load for five minutes and you guys hear me yap on. So in this, I essentially said, Hey, Claude. So again, you can create a skill conversationally. I said create a skill that scours the web and finds the latest AI news from the past 24 hours.

Let's call it the AI searcher skill. Always first take a look at the date as bringing in news for more than 24 hours ago means the skill has failed. I care about the latest AI news, LLM updates, Gen AI studies of breakthroughs and big companies, blah, blah, blah. And then I'm saying the skill should search the web and find the five biggest AI stories and give me five bullet points for each. Then it should use clawed artifacts.

So again, kind of stacking skills, and I haven't actually tried this one live. So we're going to see how it goes. And give me five bullet points. Then it should use clawed artifacts to create a beautifully designed AI-inspired news site with interactive elements that visually tell the story of the top five AI news stories of that day. Please help me make this skill. All right. So again, little meta here. It's using the creator skill to help me create a skill that then I'm going to use.

So if I kind of scroll through what happened here, I did take like three. to five minutes. When I did this right before, uh, we kicked off this show, you can go here and read, but it's, uh, creating the actual file here. Um, and it's going through, it's making some iterations. I'm only using sonnet, uh, four, five, because I know I couldn't do any demos with Opus four or five, even on a paid plan. The limits are so poor. Uh, so this obviously could have been done much better with Opus, uh, if you do have a max or an enterprise plan. Uh, so there we go. It went through and Claude actually created this skill for me. And on the right side of my screen. I actually have a preview of the markdown file and I can go in if I wanted to

and make some iterations on this. The cool thing is, and this was not available when Claude actually launched this. And I think, again, Anthropic, not always the best, if I'm being honest, at rolling out new features because usually they just roll them out to Enterprise and then like three months or, you know, their max plan. And then like three months later, they roll it out to everyone else. That was the case with skills. So they just launch skills to normal paid users on the $20 a month plan, which is what I'm on, like three weeks ago, and there's like a tweet. And that's it. So it's like now, I'm sure Claude has millions of people on their base paid plan. And I'm sure that very few people, very few of them,

unless you're like me and poke around large language models every single day, even know that this is there. Anyways, a new update that's here is now you have the ability to save this skill. right you still used to have to download it do the whole zip file upload but now you can just click and save the skill so there is obviously an advantage to using claude's built-in skill creator but like i showed you for me i love pulling in all the my data my tendencies my preferences my memory from other large language bottles and having them build skills for me based on what i do all day right so i'm going to click save skill here should just take just a second and then it says skill uploaded view. Cool. So now I can, uh, all right. There we go. And there it is. There is my AI

searcher. So let's go ahead into a new chat window. And I'm just going to say using the AI searcher skill, um, find me the top five AI news stories. So I haven't done this one yet. We're going to do this live, um, as we kind of, uh, wrap up the show here. So let me just go ahead and say this as we let Claude maybe do its thing. Hopefully this works well. But if not, don't worry about it. Claude skills are extremely powerful, right? This walkthrough was meant for beginners. This is to get you started. I wanted to show you live because, yeah, even though I've done this dozens of times, there's still little hiccups, right? I forgot, oh, you can't have capital letters in the skilled.mdfile if you're building it on your own, right? So I had to fumble around a little

bit. But I wanted to show you the basics, but you have to think of all of the capabilities of Claude, which are a lot different, I think, than Gemini, then chat GVT, then Microsoft co-pilot. And you need to think of what are those manual knowledge work tasks that you do over and over and then pairing them up with Claude's capabilities and then breaking them into multiple sets of skills. Again, don't try to automate that three-hour project. in one skill set, right? Chat with Claude, chat with chat GBT, GBT, chat with Gemini on how you can break that up and you really have to understand Claude's capabilities because like I said, great at,

great at running code, great at software development, pretty good at creative writing and reasoning and logic, but great, just on the agentic side. So we'll see here. All right, it looks like it's going through and it's at least starting to do things correctly. So it gave me the top five AI news stories. Let's see if it did a good job because we covered these, you know, yesterday. So Trump approves the invidia chips to China. That was one of our top news stories. Accenture and Anthropic launch a $200 million partnership. That wasn't a top five story, but it made our newsletter yesterday. Anthropic launches Claude Code in Slack.

That was actually a little more than 24 hours, but a lot of news organizations were covering it. you know, 24 hours ago. So heavy, heavy on the infropic side, right? And then it says Trump plans executive order blocking state AI regulations. We cover that two days ago. So okay. And then meta's AI strategy faces internal confusion. Yeah, that's talking about their new model codenamed avocado.

So did a pretty good job of all those things have been in our newsletter the last day or two. But hey, this is why you can't just. you know, hand all your AI news coverage over to AI, right? Because it did give us some older things. But you'll see what it did for a live stream audience. It went through and it, like I asked it to, because I trained it on this skill. It went and built a kind of interactive dashboard. So I have a nice kind of multicolored.

I did ask for some interactivity. There's a little bit. Yeah, the read more button doesn't do anything. But there's just some simple hover and color change effects. So it did what I asked it to. It created the kind of top five news stories, give me some information about those, and then used the artifacts feature to make some nice, nice little interactive AI news site here. So that is a wrap for today's show.

But if you're just like, whoa, this thing is like extremely powerful, but I don't know where to get started, we did it all for you. So if you want access, we built this skill library. Like I said, this is what I do all the time, right? I know all the most important use cases and have spent a lot of time aligning them with the capabilities of different large language models like Claude. So we built this skill library specifically. These are ideas, number one, right?

These aren't like copy and paste. These are skills. But you could bring these into Claude skill creator or you can bring them into the skill builder that we built. So if you want access to this database, this skill. library database and it is a database it is filterable sortable by different categories so uh you know maybe you're in hr you're in sales um you're in marketing etc we have different uh skill library ideas uh that you can sort by category and you can also sort by level beginner intermediate and

advance so if you want access to that skill library and uh the skill builder gb t or the skill builder gem all you got to do is just repost this show on LinkedIn. So this is something I'm trying to do more of because everyone's like, oh, Jordan, how can I help? And I'm like, tell someone about this, right? So if this podcast is helpful, I spend a ton of time, right? Y'all, I'm sick, right? I'm struggling to talk, right?

If you really want to help us and support the show, go repost this show and I'll send you these things. But I'd really just appreciate it if you just tell people, right? This AI stuff, it's so hard to keep up with. And as we prepare for 2026, I think having an unbiased source that you can trust that can break things down in a simple way in 20 or 42 minutes is extremely important. That's what we try to do. And we cannot keep this thing going without your support.

So please go repost this exact show on LinkedIn. And I will send you those resources. If you are listening on the podcast, we always put those links in the show notes. So make sure to go check it out. So thank you for tuning in. I hope this was helpful. Remember, go to our website, Your EverydayAI.com. Check out today's daily newsletter.

We're going to be recapping today's show and a whole lot more. So thanks for tuning in. We'll see you back tomorrow and every day for more Everyday AI. Thanks y'all. Meet Firefly AI Assistant. Now live in Adobe Firefly, the Allman One Creative AI Studio. Just describe what you want to create in your own words and the assistant handles the rest, orchestrating multi-step workflows across Adobe Creative Cloud apps,

including Photoshop, Premiere Express, and more in one conversational interoperational interface. You direct the outcome while the assistant accelerates execution. Stand control with the ability to step in and refine at any time. See it today at firefly.adobie.com. And that's a wrap for today's edition of Everyday AI. Thanks for joining us. If you enjoyed this episode, please subscribe and leave us a rating. It helps keep us going. For a little more AI magic, visit Your EverydayAI.com and sign up to our daily newsletter so you don't get left behind. break some barriers and we'll see you next time.