9 Free AI Skills That Feel Like Cheat Codes

One of the best things about tools like OpenAI's Codex and Claude Code or Claude Cowork is that there are a lot of people using them already and figuring out the best way to actually get results from them. And many of those people, instead of, you know, like hoarding that knowledge to themselves, actually go and make their findings available for anyone to duplicate. And they do that in the form of what we call skills and plugins. And people are using these skills to do crazy things like these awesome knowledge graphs and to do deep real-time research and to act their CEO and to even create animations for them that they could use inside of videos. So in this video, I'm going to share my absolute favorite skills and plugins that I've come across that are completely free to install in Codex or Claude Code or whatever your harness of choice is. These skills and plugins have become pretty universal, so they'll work in like any of the IDEs or agent harnesses that you use, you know, Codex, Claude Code, Cursor, OpenClaw, Hermes, VS Code, Microsoft Copilot, like all of those. Now before I get into my favorites, let me quickly explain what skills and plugins are. If you wanna skip over the explanation, just jump to this timestamp to dive right into my favorites. All right, so a skill is basically a reusable instruction file. It's typically just like a detailed prompt written inside of a skill.md file. There's also sometimes some additional references, scripts, examples, or assets to go along with it, but most of the time it's essentially just a prompt prompt that's already been dialed in so that you get a consistent result whenever that skill is called upon. Then when you go to write your own prompt, the large language model reads both that skill file to help it understand what to do and the prompt that you just gave it, so that both are fed into the model at the same time. A plugin, on the other hand, is more of like a bundle. It could package skills, agents, hooks, MCP servers, commands, or other config so that someone else can install the whole thing instead of like manually copying folders. So a skill is essentially like the single prompt that tells the model what to do whenever that skill is called upon. A plugin is much more of a bundle that has not only skills but agents and MCPs and all sorts of other stuff. But the magic in all of these is that they don't just give your agent more context, they kind of give it like a repeatable behavior because every time you call on one of these skills, or plugins, it knows how to act every single time. All right. Now that I've explained that poorly, and I fully expect the comments to tell me what I got wrong, let's dive into my favorites because let's be honest, that's what you came for. You're not here to learn. You want to see this do some cool stuff. The first one I want to talk about is called GStack, and it comes from Gary Tan, who's the CEO of Y Combinator. Now I'm not going to get too hung up on whether or not these different projects I'm talking about are skills or plugins. At the end of the day, you're going to set them up pretty much the same way. This one is essentially a giant bundle of a whole bunch of skills that you can give your agent. Here's how Gary describes it. It turns Claude Code, but it also works in all the other things as well, into a virtual engineering team, a CEO who rethinks the product, an engineering manager who locks architecture, a designer who catches AI slop, a reviewer who finds production bugs, a QA lead who opens a real browser, a security offer who runs OWASP and STRIDE audits. I don't even know what that means. And a release engineer who ships the PR. There's 23 specialists and 8 power tools, which are all slash commands, all markdowns, and all free to use. And we're going to install most of these tools pretty much the same way. Here's how simple it can be. I'm going to grab this GitHub URL up here, github.com/garrettan/gstack. And if I was installing it in the Claude Code app here, I'm simply going to come down to my prompt box and say, install this for me, please, and then just give it the URL to the GitHub. Literally, that's how easy it is. If I'm doing it in Codex, I could do it exactly the same way. Now, one caveat with Codex is they do have this plugins and skills marketplace. So some of the plugins and skills that we're going to look at, you don't even need to have it installed that way. You could just go and click this add button next to it and it'll just be in Codex. But if I search GStack, you'll notice that it's not one of the available tools in here. So I'm going to come down to chats down here in Codex and say install this for me, and then paste that GitHub repo right in there. Let it go to town installing that for me. And we can see we've got it now installed in Claude. And if I jump over here to Codex, it's installed here in Codex as well. I'm pretty much going to follow that same process for every skill or plugin that I install, but just know that all of the plugins and skills that I show have been shown to work in Codex, Cowork, Claude Code, OpenClaw, Hermes, Visual Studio Cloud Code, GitHub Copilot, they work across the board in all of them. Just pretty much follow that same simple process, just ask the agent harness to do it for you. So for the remainder of this video, I'm just going to work in Codex to simplify things and keep the video moving along. Now, just to show you that this did install in Cloud Code and is working, I'm going to create a new session. I'm actually going to open up my little test folder here, and I'm going to give it the command /gstack office hours. Office hours is one of the skills that's built in. I do have to give it this /gstack the way we installed it here. And then I'll say, I want to build a web app that takes a YouTube channel URL, analyzes the last 20 uploads, and suggests 10 new video ideas. Help me pressure test the idea before we build. I'll submit that, and it's going to call upon our gstack office hours skill here, and it's going to stress test this idea. Now, while that's running, I'll jump back over to Codex here, and we can see that if I come up to plugins, click on skills at the top, All of these GStack skills are now installed. And again, this is a giant bundle of various skills. I'm not going to go through all of them because, well, there's a lot. But I'll open this business dashboard project that I have here. This is a dashboard that I created that tracks all of my various social stats across platforms for me. And I'll give it the prompt gstack-review. I'm going to use the review skill and say, review the current changes like a senior engineer. Look for bugs, missed edge cases, security issues, and UX problems. So it's using that GStack review skill and it's going to go ahead and review the code for me here. Jumping back over to Claude Code to see how we're doing on my idea for a idea generator, it's starting to ask me some questions here. What is your goal with this? Startup slash real product, internal entrepreneurship, side project, hackathon slash demo. For this, let's go side project. Over on the GitHub repo here, Gary suggests doing these 4 skills first. Run the office hours skill and describe what you're building. Run the plan CEO review on any feature idea. Run review on any branch with changes, run QA on your staging URL, and then stop there and you'll know if this is for you. But again, there is a ton of skills in here, so just explore and play around with these various skills when you're writing code or planning projects because there's a lot you can do here. Jumping back to the Codex version here, it found one issue with a recommended fix, and it's actually a pretty helpful fix because I changed the scope of this project over time, but some of the old scope is still built in. So something I probably need to fix. I'm still getting more questions over here in my Claude Code about my idea because it's trying to stress test it for me. It's giving me a lot of questions that are helping me stress test with the AI. So I'll keep answering the questions in the way that I think this app should be built. And this wasn't actually an app I was planning on building, but who knows? By the end of this process, maybe it's something I'll build. All right. So in Claude Code, this stress test took about 20 minutes, but it went back and forth and back and forth and asked me a ton of questions so that I can stress test with it to decide if the idea is actually good or not. It wrote up a brief for me that it put into a markdown file and really just sort of helped validate the idea. But again, that's just scratching the surface of all the skills available here. So again, go play with it. Now the next skill I want to share is pretty easy to explain. It's called Stop Slop, and it's a skill file for removing AI tells, basically. Once again, simple install. I will grab this GitHub file, jump into my Codex here, create a new chat, install this for me. This one is very pretty small compared to GStack, so it should be pretty quick. While it's doing that, I'm gonna have ChatGPT write me up just a one-paragraph explanation here about humanoid robots. I'll copy this one-paragraph explanation here. It's telling me to restart Codex, so I'll go ahead and do that. And if I look in my plugins skills, my Stop Slop skill is installed now. So let's create a new chat here. I will give it the prompt, use the Stop Stop Slop skill to improve this paragraph. And I'll go ahead and paste in the paragraph we just got from ChatGPT, and it's going to use the Stop Slop skill. And here's our new paragraph. I'm not going to read them both out loud, but if you do want to pause and see what it did, feel free to do so. If you do any AI writing like scriptwriting or blog writing or anything like that and you want to remove some of the AI-isms from it, this is a really simple skill that'll do that for you. So the next skill I want to get into is actually one of my favorites. I really love this skill. It's called Graphify. Now Graphify turns codebases and docs and schemas and notes and media into like a queryable knowledge graph, and it makes it so that your agent or, you know, your coding harness or whatever you're working with doesn't have to rediscover all of the same relationships over and over again. It can actually just query the graph like memory. So it's actually got this really cool visual workflow, but also it's got this memory component that will save you on token costs because it can use this as an additional memory. So we'll pop over to the GitHub page for this one. All of the links to the skills I'm talking about will be in the description. We'll copy this URL, jump on into Codex. I'm gonna start a new chat down here and I'm going to say, install this for me. And just to save time for the rest of this video, I'm gonna do this same process to install all of these skills. So I don't need to keep showing you this, I don't think. Now one really cool use of this Graphify skill is if you've been building like a second brain. If you watched my previous video where I built like a whole knowledge base around content that I come across online, it's really good at sort of mapping that all out for you. So let's go to my second brain project here. I'm going to create a new chat and I literally just have to type /graphify and then a period, and it's going to graphify this project in this folder for me. You can see I'm in my second brain folder. It asked me which part of my second brain I want to Graphify. I'm going to tell it the wiki and journal. Those are the ones I wanted to Graphify. Okay, so it took about 6 minutes, but it created our graph. And check this out, if I open up our graph.html file, we have this node-looking graph that looks similar to what you'd get in the graph in Obsidian. Here it is in Chrome to get a better view of it. But if we zoom in here, we can see all of the interconnections of various things I've saved into my wiki. And notice some patterns, you know, YouTube Creator Strategy, Creator Sustainability, Creator Persistence. Down here we have AI Agent Workflows, Second Brain Usefulness Plan, and a bunch of random like non-connected nodes, but it makes for a pretty cool visual. Now these communities over here are sort of like bundlings that it found. So if I unselect all of them, it makes the whole thing disappear. But let's say, let's say I want to just see the bundle of YouTube Creator Strategies. I can select this right here and zoom in and we can see some of this stuff I've saved around YouTube. Creator strategy. I can click on any section, see some of the neighbors. Like if I click on social media algorithms, neighbors are AI Slop, audience avatar, information gain, etc. So you get this nice little visual of all of the stuff you've been saving into your database here. But again, this is just a cool visual. What makes this really helpful is that it's actually this new memory layer. Like it helps your agent see the interconnectedness of stuff so you can actually query it. So if I close out of this here, I can ask something like, using the graph, what are the biggest recurring themes in my notes? And you can see it says, I'll use the existing graph rather than rebuilding it. So it's going to query the merged graph JSON file and we can see it found some connections. AI workflows as judgment leverage, creator strategy under platform pressure, persistence, originality and creative independence, behavior design and making hard work easier to start, sustainability as a creator, the second brain, AI era positioning, etc. So it found all these connections of all of the things that I've saved, as well as the things that I've journaled on into my second brain. Let's ask, what are 3 video ideas hiding in this knowledge base? It says, I'm going to look for high-degree bridge nodes and thin but interesting clusters where the graph suggests an underexploited connection. And it found 3 based on things that I've saved. AI won't replace creators, it will expose their judgment. The creator trap: when YouTube strategy starts eating your originality. The new 4-Hour Workweek is not about working less, it's about building with leverage. It also gave some details on, you know, the graph connections and the angle for the video as well as the hook. But the nice thing about this is when I ask these questions, it's not trying to query across the entire database of all the markdown files of everything that I saved in my second brain. It's literally just looking at the graph that it just created and pulling these ideas from just that, which is going to use way less tokens and be way more efficient. Graphify also works on codebases. I just ran it on my Future Tools codebase as well. And you can see it built this graph here of Future Tools, a little bit more complex than you probably realize. But again, using this as the memory layer is the real power here. It even suggests some questions to ask here, like why does CN connect so many UI communities? So if I ask it that, it's going to actually query the graph and give me the answer without digging through all of the code files individually. And here's my answer. Now, the next skill I want to talk about is similar but also different. It's called Understand Anything, and they describe it as graphs that teach are greater than graphs that impress. Turn any code into an interactive knowledge graph you can explore, search, and ask questions about. So this one is slightly less about memories for the agent and more about like interactive onboarding and an exploration layer for humans and agents. So imagine this scenario. You write this big codebase on some project and then you onboard a new hire to work on the same code. You have them run something like Understand Anything on the code and it essentially onboards them to what's going on inside of the code. But it does still create some pretty cool visuals. So for example, if I want to have Understand Anything map out everything that's going on inside of the Future Tools website, I can come to my Future Tools project here in Codex, type /understand, hit enter, and it will run the Understand Anything skill so that it will map out everything that's going on on the website. After almost 9 minutes, we have it all done here. And this is what this one looks like. A little bit more detailed of the flow of how all the code is working together. We've got our API routes, our security boundary. This is flowing into our services and integrations. So is our Supabase data model, our admin workflows. This is all funneling into the public UI. We have documentation, project configuration,. It's just a very visual workflow of how this site is built. A lot easier to grasp than that giant cloud we were just looking at with Graphify a minute ago. And I can click into any one of these to dig a little bit deeper. So if I click into services and integrations, it will show me a map of all the services and integrations and how they all connect together. So, you know, you've got your news, which feeds into server and also feeds into types. We've got mock tools, which were originally just sort of like demos that I put in that probably don't need to still be there. Come up here. We can see index and pipeline and Firecrawl and Anthropic all sort of tie together and how Google Sheets and Beehive are tying into our API routes. And it's just a much more visual flow so that somebody that's a coder could kind of come in here and see how all of the pieces fit together. Now, because I'm a vibe coder and I don't totally know what I'm doing, it's probably a giant spaghetti mess. Somebody that actually knows what they're doing, it would probably be a lot cleaner looking of a workflow, for being honest here. But if I jump back to Codex, I can ask it things like, where should a new developer start on this? And it's actually going to look at the workflow again as sort of like a memory to help answer that question, as opposed to digging through all of the entire code again. And it's going to give me a quick explanation of where the developer should start. And here's what Understand Anything looks like when I run it on my second brain. We have a giant knowledge map of how everything interconnects. A little bit more hectic for this second brain concept here, but that's a view of all of the content and journal entries I've saved into my second brain. If I click into one of these, like this Riley Brown video, you can see that it shows the route of how these connect to each other. So it's connected to my index page down here. It's connected to probably another Riley Brown video and another Riley Brown video. So you can click on any one of these and see how they connect together with other sources in here. Let's just click on another one at random and we can see how they all connect together. You know, not quite as useful as using this for code, especially when your knowledge base gets very large, but still a really cool visual to click around and play with. All right, before I move on to the next skill, I want to shout out really quickly, if you haven't already, make sure you subscribe to the Future Tools newsletter. Head on over to futuretools.io/newsletter and join. We have over 250,000 subscribers now. And every single week I'm going to send you 2 emails per week with the most important news that you need to know in the AI world, as well as the coolest tools that I came across that week. It's totally free to subscribe. You also get some extra goodies like my AI Income Database, which is a database of all sorts of ways that people have found to make money using AI tools. It's like a make money life hacks and you get it for free when you sign up to the newsletter again over at futuretools.io/newsletter. All right, let's move on to the next skill. The next skill I want to talk about is called Last 30 Days. And this is a great skill if you're doing research or you want to find out what like the internet sentiment is on a specific topic. So what this one does is it helps you figure out what people are actually saying across the internet right now. It searches things like Reddit, X, YouTube, Hacker News, Polymarket, GitHub, and just the web in general. And then it synthesizes a grounded summary on what it finds. So I went ahead and installed this one in Codex already using the exact same method we've done in the past. And if I click on chat here to start a new chat, I can type /last30days to call upon this skill. Just for fun, let's do research Claude Fable, find what people are excited about, what they're confused about, and what they're speculating will happen. And it gives me back an explanation of the sentiment around Fable right now. People are excited because Fable feels like a real capability jump. Developers are excited about coding agent leverage. Confusion is mostly about access, citizenship, and shutdown mechanics. People are confused by Fableverse mythos. And it's also sharing all the sources where it found it, like it's looking at Reddit and YouTube, Anthropic's announcement, the system card, things like that to pull in details. At the bottom, it gives us the key patterns. Excitement centers on work completion. Availability drama drowned out launch hype. Cost is treated as destiny. Safety classifiers are the technical mystery box. The next bet is restoration plus workarounds. And then it tells us what what agents actually reported back. So it pulled 21 threads from Reddit, 20 videos from YouTube, 27 stories from Hacker News, 19 items from GitHub, 4 markets on PolyMarket. And it goes on to say, I'm now an expert on Claude Fable 5. Some things I can help with. And then it gives you some follow-up questions you can ask. Another really cool thing is you can say, use the same skill to generate a shareable HTML file, and it actually it has instructions inside of the skill to design out this HTML file. You can see it says, I'll use last 30 days HTML export path and reuse the saved Claude Fable research. So it already has an HTML export, it already knows what to do with that information. And here's the HTML page that it created. Pretty much the same information, just, uh, you know, easier way to view it and something that you can share with others. All right, the next two skills I'm going to share are kind of similar. So I'm just going to sort of lump them together. One of them is the Frontend Design skill, which is actually made by Anthropic. And if we actually look at the skills.sh website, which is a site that helps organize and find skills, it's actually the second most popular skill of all time here. And this one does pretty much exactly what it sounds like. When you're designing a web page or an app, it will help with the aesthetics of the frontend design. Now, the other one that I want to talk about is called the Taste skill. This one comes from Leon XLNX and it's got nearly 50,000 stars on GitHub, but pretty much does the same thing. It gives your AI good taste, stops AI from generating boring generic slop specifically for your front end. So if you're building a website or an app or something like that, again, this is going to help with the design of it. I've installed both of these skills on Codex and I asked each skill to redesign the Future Tools homepage. So for a quick reference, this is what the Future Tools homepage looks like right now. Out, right? It's the AI Tools database. It's got all the categories you can select from. You can search for AI tools, and it's got, you know, a giant filtering of all the AI tools that are available. That's our starting point. For my first test, I asked it to look at the current homepage design, redesign it so it looks cleaner and more eye-pleasing, and create a new HTML page so we can see what it looks like. But for this first test, I said don't use any additional skills to accomplish this, just use the model by itself. Here's what it designed for the new Future Tools website when it just designed it with no design skills whatsoever. Again, here's my current site. Here's the new one that it just designed. Pretty much fairly similar. I do kind of like how it put the call to action off to the right instead of being right here. That's a little bit cleaner than what I've got, honestly. I then gave the exact same prompt, but this time said, use the frontend design skill to accomplish this. And here's the design using the frontend design skill. Definitely more colorful. Again, it kind of did this similar thing where there's the top is sort of split. Definitely a better design than this really basic design that the model just did by itself. So my site, the model without any design skill using Anthropic's design skill, here's what it came up with. I then give it the exact same prompt but told it to use the Design Taste front end, and this is what it designed using that one, which is a much cleaner look. But it sort of removes what the existing website is designed for, which is, you know, a curated list of all of the tools. So good design just, you know, removed all of the tools from the design. And then for my final test, I told it to use both the Design Taste front-end skill and the front-end design skill. So combine the two skills into a single prompt and use them both. And here's what that generated. This has got a little bit of a gradient here. I don't like how big this image is over on the right. That's a little weird. And I mean, it looks decent,, but those are the examples. Obviously aesthetic is a very subjective topic, so you decide. This is my original. This is very plain, boring. I kind of designed it to be like a Craigslist kind of very simple site. Here's what AI designed on its own. Here's what it designed when given the Anthropic front-end design skill. Here's what it designed when given just the taste skill. And then here's what it designed when told to work with both of those skills. Probably my least favorite design of all of 'em, quite honestly. But those are two skills that you can use if the design you're trying to get for your site or your app aren't quite coming out the way you want them to by just using the model alone. You can install one or both of these skills and get multiple variations of your design to get something that you really, really like. There's two more skills I want to show off, and once again, these two skills are very, very similarly related. One of them is called Remotion, and the other one is called HyperFrames. And both of these are skills that are designed to actually use your coding app to generate animations. ReMotion is kind of its own thing, and HyperFrames is made by HeyGen here. But you give each one of them a prompt for a video you'd like to see it generate, and then it will make sort of After Effects-y style videos. Not quite as good as what you can do in After Effects. If you want something really, really polished, you're still probably going to want to hire like a visual designer or somebody that knows After Effects. Effects, but it'll do really decent, like, basic graphics if you just need charts or you need just text to animate on the screen, things like that. These are really good at that kind of thing. So for my first test here, I wanted to test a ReMotion, so I used the ReMotion best practices skill and I said create an animation that looks like a text conversation on an iPhone. And then I gave it a quick text dialog to generate. It generated this MP4 that I can then view on my computer here. So it's an iPhone. Have you subscribed to Future Tools No. Why? Because it's literally how people are keeping up with AI news these days. Oh, okay. That's how they're doing it. Just subbed. Thanks. I mean, that was one prompt and it made that animation for me. It made this 11-second animation of a text conversation that I can now use in a video. I then gave Codex the exact same prompt, but this time called upon the HyperFrames skill to do the same thing. And here's what that one looked like. Quite a bit better. Honestly, I would say that actually looks more like an iPhone than the other one. I don't like how the text on the blue bubble is black, but other than that, it did a really, really good job of making this little iPhone animation from just one prompt. That was just one prompt through. Both of them did a pretty good job. So if you're trying to get this kind of animation, you can use both of those skills and see which one does it better. I wanted to test a logo reveal, so I gave it the Future Tools logo here and said create an animate— I didn't even spell it right— create an animated logo reveal It starts where the logo is not showing, then there's a particle explosion, and then the attached logo is formed from the particles that exploded. Use the ReMotion best practice skill. So this is the version that ReMotion created. Nothing, particle explosion, and then they all form into the Future Tool logo just like that. Pretty good. So if you need like a logo reveal animation, you can now get AI to code it up for you just like that. And of course I tested the same animation with HyperFrames, identical prompt except I swapped out ReMotion with use the HyperFrames skill. And here's how HyperFrames did it. We have a blank screen, an explosion, and then it forms into the Future Tools logo for us. Pretty good. I think I like that second one a little bit better, actually. I mentioned it was good at charts and graphs, so I want to do another example here. Create an animated stock chart that shows the growth of NVIDIA stock over the last 5 years. Here's how Remotion did it. We could see our stock chart, and little by little we can see how the NVIDIA stock sort of exploded over the last 5 years. And it even used Yahoo Finance to get the source of that information. And here's the version that HyperFrames generated. A little bit more animation of things going on inside this video. Definitely think the HyperFrames one is the preferable version here. And that's what I got for you today. Some of the coolest skills and plugins that you can plug into Claude Cowork, OpenAI's Codex, or any agent or harness that you're using to actually write your code and do this sort of thing. More and more of these are coming out all of the time. So if you want to see more breakdowns of cool skills and plugins that you can use directly inside tools like this, let me know in the comments and I will definitely make more videos like this if there's enough interest and demand for them. Again, that's what I got for you. If you like tool breakdowns and tutorials like this, make sure you like and subscribe to this channel and I will make sure more stuff like this shows up in your YouTube feed. Also, every single Friday I make a news breakdown video where I share everything that happened in the world of AI that I think the most people need to know about. They come come out pretty much every single Friday without fail to help people stay looped in on what's going on in the world of AI. If that's your thing, also make sure to like and subscribe because those will show up in your feed as well. Anyway, that's what I got for you. Thanks for hanging out. Thanks for nerding out with me. Really, really appreciate you, and hopefully I'll see you in the next one. Bye-bye.