The argument in one line.
When everyone has access to the same AI models, the differentiator is not the tool but the skill—how you prompt, structure workflows, and layer capabilities like visual generation, scope-mapping, consensus-building, and devil's advocacy to extract maximum value.
Read if. Skip if.
READ IF YOU ARE…
- You're a content creator or educator who regularly explains concepts visually and wants to build interactive infographics or diagrams without design skills.
- A solopreneur or small team using Claude daily who wants to work faster by discovering and automating your most repetitive prompt patterns.
- You're experimenting with AI workflows and want concrete examples of how to layer skills (prompting, visualization, analysis) to produce better outputs than default chat.
SKIP IF…
- You don't use Claude or prefer other AI models — this is entirely Claude-specific and won't transfer to ChatGPT, Gemini, or other platforms.
- You're looking for production-grade design or engineering solutions — these are productivity skills for solo creators, not enterprise tools or replacements for professional designers.
The full version, fast.
When everyone uses the same models, the skill layer is what separates outputs. The video walks through seven daily-driver Claude skills, splitting them into chat-based and Claude Code categories. Infographic Builder and Excalidraw Diagram Generator turn ideas into interactive HTML visuals or whiteboard flowcharts. Expand & Contract maps an idea's full scope, then sorts features into core, nice-to-have, maybe-later with triggers, and explicitly out. Steel Man stress-tests a thesis with arguments for and against. Promptimizer rewrites vague requests into structured prompts. Swarm Consensus, run in Claude Code via an OpenRouter API key, fans a question across multiple frontier or cheap models and synthesizes agreement. Find Skills queries skills.sh's library so you stop rebuilding from scratch. Stack them, and tell Claude to flag repeatable workflows as future skills.
Members feature
Chat with this breakdown.
Modern Creator members can chat with any breakdown — ask for the hook, quote a framework, find the exact transcript moment. Unlocks at T2: refer 3 friends + add your own API key.
Create a free account →Where the time goes.
00:00 – 00:44
01 · Intro — skills as differentiator
Opens on infographic output, delivers thesis: same models + different skills = different results. Frames video as a curated list from real daily use.
00:44 – 04:22
02 · Skill #1 — Infographic Builder
Builds an interactive HTML infographic for 'selling digital products on Etsy' live. Skill reads types.md and aesthetics.md to pick layout and style. Output is embeddable and editable.
04:22 – 07:13
03 · Skill #2 — Excalidraw Diagram Generator
Generates flowcharts and wireframes in Excalidraw's sketchy format from a text description. Free to use. Demo: listing-publish flow and skill-install flow diagram.
07:13 – 12:35
04 · Skill #3 — Expand & Contract
Takes an idea, expands to 25+ possible features, then batch-sorts them into Core / Nice-to-Have / Maybe Later (with triggers) / Explicitly Out. Outputs a concentric-circle scope map.
12:35 – 15:42
05 · Skill #4 — Steel Man
Evaluates a statement by building the strongest case for AND against each angle, then scores whether you're winning. Demo: fitness app to $50K/month — Alek loses most arguments.
15:42 – 17:05
06 · Skill #5 — Promptimizer
Turns a vague prompt request into a refined, detailed prompt ready to use. Works for chat, Midjourney, or any AI tool. Demo shows Midjourney cyberpunk prompt with noticeably better output.
17:05 – 22:18
07 · Skill #6 — Swarm Consensus (Claude Code)
Requires Claude Code + OpenRouter API key. Sends any question to 8 models in parallel, synthesizes consensus, highlights disagreements. Creative query: 3 cents. Legal/high-stakes: auto-upgrades to frontier models.
22:18 – 24:08
08 · Skill #7 — Find Skills
Claude Code skill that searches skills.sh (90,000+ community skills) from within a chat. Reduces having to build from scratch. Closing note: add one line to CLAUDE.md to have Claude proactively suggest skill creation.
Lines worth screenshotting.
- In a world where everyone has the same AI models, skills are what determine who gets the most out of them.
- The infographic builder skill generates an interactive HTML infographic — not a static image — which means you can edit colors, text, and animations after the first draft.
- Five skill types for different visual needs: timeline, comparison, stats hero, editorial, and bold-pop — matching the type to the content is as important as the prompt.
- The Excalidraw skill generates a JSON diagram file that opens directly in Excalidraw for further editing — it is a wireframe tool, not a final visual.
- The grill-me skill resolves every decision branch downstream of a design choice before writing code — it eliminates the hidden assumptions that break production builds.
- The caveman skill cuts token usage by up to 75% by stripping fillers, articles, and pleasantries while preserving technical accuracy.
- The expand-and-contract skill maps the full scope of a topic first and then contracts it to the specific slice that matters — useful for avoiding scope creep.
- The steel man skill builds the strongest possible version of an opposing argument before you decide whether to proceed with your original position.
- The Promptimizer skill rewrites your prompt to extract better output from the same model — the prompt is often the bottleneck, not the model.
- The swarm consensus skill runs the same prompt across multiple models via OpenRouter and returns a synthesized answer weighted by agreement.
- The skills.sh marketplace provides a curated directory of downloadable Claude Code skills — discovering what exists is the first prerequisite to using what works.
- Knowing when to use a skill matters as much as knowing how — applying the right skill to the right bottleneck is the judgment that separates good output from average output.
The skill layer is the moat.
Format to steal
Every skill Alek demos follows the same loop: invoke, configure, output, so-what — and that loop is itself a content format Joe can run weekly on JoeFlow skills.
- Run a '7 skills I use daily' format for JoeFlow — each skill gets 2-3 min of demo + one real output shown.
- Lead with the OUTPUT frame (infographic on screen before you talk), not your face — it front-loads proof.
- The thesis sentence ('same models, different skills') is Joe's 'own your stack' translated to AI workflows — use it verbatim in JoeFlow positioning.
- Expand & Contract is worth installing in Claude Code today — direct fit for scoping Mod Producer runsheets.
- The one-line CLAUDE.md instruction ('note opportunities for skill creation while working') is a free compound-interest trick — add it to the universal CLAUDE.md.
- Swarm Consensus + Steel Man together = an idea-validation pipeline that would make a strong MCN+ onboarding flow.
Terms worth knowing.
- Claude skill
- A reusable instruction package that extends an AI assistant with a specific workflow, invoked by name to make the model behave a defined way for a task.
- Infographic builder
- A skill that generates an HTML-based visual instead of a flat image, letting the output include tooltips, animations, and editable elements rather than a baked-in picture.
- Excalidraw
- A free web-based whiteboard tool for sketching diagrams and flowcharts in a hand-drawn style, with a file format that can be imported and edited later.
- Expand and contract
- A scope-setting skill that first brainstorms every possible feature for an idea, then forces the user to sort each item into core, nice-to-have, maybe-later, or explicitly out.
- Steel man
- An argument technique that builds the strongest possible version of a position before critiquing it, used here as a skill that argues both for and against an idea to stress-test it.
- Devil's advocate
- Deliberately arguing the opposite side of a position to expose weaknesses, regardless of what the arguer actually believes.
- Promptimizer
- A skill that rewrites a rough request into a fuller, better-structured prompt, asking clarifying questions first when the input is too vague.
- Midjourney
- A generative AI service that turns text prompts into images, commonly used for stylized art and design work.
- Claude Code
- Anthropic's command-line coding environment for Claude, where skills can be installed into a project folder and invoked alongside file edits and shell commands.
- Swarm consensus
- A skill that sends the same question to several AI models in parallel, then synthesizes their answers into a single response highlighting agreement and disagreement.
- OpenRouter
- A routing service that gives a single API key access to most major AI models, so one integration can call Claude, GPT, Gemini, Grok, and others.
- API key
- A secret token that authenticates a program when it calls an external service, and is also used to track usage and bill the account that owns it.
- Frontier models
- The most capable current generation of AI models from major labs, typically slower and more expensive than smaller models but stronger at complex reasoning.
- Claude Opus
- Anthropic's largest and most capable Claude model tier, priced higher per token and used for harder reasoning tasks.
- Claude Sonnet
- Anthropic's mid-tier Claude model that balances capability and cost, suitable for most everyday tasks.
- Claude Haiku
- Anthropic's smallest and fastest Claude model, optimized for cheap high-volume calls where top-tier reasoning isn't required.
- Grok
- An AI chat model developed by xAI, Elon Musk's AI company, available through its own product and via routing services.
- ICP
- Ideal Customer Profile — a specific description of the type of buyer a product is built for, used to focus positioning and marketing.
- Churn
- The rate at which paying customers cancel a subscription over a given period, a key health metric for any recurring-revenue business.
- SaaS
- Software as a Service — software delivered over the internet on a subscription, where the vendor hosts and maintains the application.
Things they pointed at.
Lines you could clip.
01:00
“In a world where everyone has access to the exact same AI models, skills determine who gets the most out of those models.”
19:38
“I've done around 30 sessions with the most intelligent models, and it's only cost me $15.”
23:03
“While working note opportunities for automation, improvement, repeatability, and if a task is a good candidate for a Claude skill, tell me so that I can turn it into a skill and reuse that workflow later.”
Word for word.
metaphoranalogy
00:00So around 65% of people are visual learners, and that's why you might need some kind of infographic or visual to help explain your point. And, of course, you also get some bonus points if it's interactive.
00:10This is the output from our first skill, the infographic builder, and has a pretty wide application from making social media content to teaching something or just building out interesting and interactive visuals. And I'm gonna show you exactly how it works in just a second, but in a world where everyone has access to the exact same AI models, skills determine who gets the most out of those models and knowing how to use them, when to use them, and how to make them is a skill in and of itself.
00:36Today I'm gonna share with you seven skills that I use almost every single day. I'll explain how they save me loads of time or help me make money, and I'll give you a demo of each one. So this visual that you're seeing right here is an infographic that I built with the infographic builder skill, and it allows you to build out visuals that are either interactive or just look good to help you explain a point in just a few minutes.
00:57This is different from an actual image because it basically builds out an HTML page with all of the elements that you want in that infographic or diagram or whatever visual you're trying to create, and it can also make them interactive. So if you want to hover over some text that can display a tooltip, you can highlight the text, you can add different animations for loading for example, and it can be a bit responsive depending on the page size where you're displaying it.
01:22I like making infographics this way because it could still be used as some form of social media image post if we grabbed a screenshot of it, but we can also edit things like the colors or the specific text or even the animations that are used in this infographic better than if we just tried to create a static image. So we'll make a new one here in a second, but if I open up the skill file, you'll be able to see sort of how this works.
01:44So first, it's gonna run an intake. So it's going to kind of try and understand what type of infographic you're actually trying to build. They'll confirm it.
01:52It'll pick a type. So there's this other types markdown file where we've laid out a few different types of infographics depending on what needs to be displayed. So we have a timeline view, a comparison view, a stats hero, and so there's a few different types of infographics that it can actually make depending on the requirement from the user.
02:13Then it's gonna go through and pick an aesthetic. And so it'll reference the aesthetics markdown file and you can see here there's just a bunch of different styles that it can choose from. So we've got like a minimalist style.
02:24We've got editorial style, bold pop, and so there's a few different looks that the infographic can actually have. And from there it'll lock in the dimensions. It'll question you about how interactive you want it to be.
02:35So if you want some of the things like tool tips or maybe clickable elements within that infographic, then it'll ask you if you want those and how many of those you want. It'll build it, review it, and then you'll have your infographic. But now we can actually use it together.
02:48So I'm gonna start a new chat. And in here, I'm going to invoke the infographic builder skill. You just type a forward slash and then type out the skill name that you want to invoke.
02:58And then from here, can tell it what we want to build out. So I might say, make me a quick Instagram post infographic explaining how to sell digital products on Etsy.
03:07Here we get a few questions to answer. I think we'll go with a square post. And what visual direction.
03:12Let's go with the soft organic look and what content angle. What's the hook? We'll go with stats driven for example.
03:20I like that it prompts you with questions like this instead of just assuming exactly what you want. It oftentimes gets you a better result on the first try. If you have actual data to plug into these infographics, it'll work a little better.
03:30Right now, Claude's having to go do the research and then use this skill to build out the infographic. But after just a few minutes, there we go. Now we have a full infographic about selling digital products on Etsy.
03:40Honestly, this looks incredible in my opinion. Uh, make once sell forever.
03:45So this is specifically an infographic tailored around selling digital products. But if you're trying to make a carousel on Instagram or some posts for threads or you just want a visual to teach or explain something, something like this makes it incredibly easy to create good looking visuals that you can come in, you can edit, add animations to, essentially just build out high quality infographics depending on what you need.
04:08The only thing that this one missed is that it's not a square, so we could come back and have it rerun and try and fit everything into a square if we need to. Whenever I go to make visuals and I need something that's just a little bit better than a static image, this is what I typically use, and I think the results are pretty good.
04:23But that brings us to the next skill on the list which integrates with this website called Excalidraw. Now you've probably seen this before if you've seen any other videos talking through complex topics or just showing off sort of a wireframe or visual in this sort of scribbly whiteboard layout.
04:40But instead of explaining it, let me just show you how it works. So if we open up a new chat and I invoke the XRAW diagram generator, then we could tell it something like, explain the process of posting a listing view listing from listing view to Etsy following this process. And then I'm just going to paste in the process that it should follow, and we're gonna hit enter.
04:59This one's cool because right from the beginning, it's gonna ask you if you wanna customize it or just auto generate. For this one, we'll just auto generate with whatever style it feels will look the best. This is another sort of visual skill, but it's really helpful when you want to teach or show something, and we can do another example.
05:15So I'm gonna say, use the X Calc diagram generator skill and build me out a diagram explaining how to download and install skills directly into Claude. Alright.
05:26This first one is done. So I can just hit open, and we need to download it to our computer.
05:31By the way, you can use this completely for free. So if you come to their pricing page, you can just click on draw now on the free account and it'll open up a whiteboard where you can start to play with all of the elements.
05:41You can control all of the text boxes, all of the text inside the text boxes. And if you come over to the sidebar, we can just open, and then I'm gonna open up that listing view publish flow, and you can see that we have our diagram prefilled right here. So starting a listing optimizer, we create a listing.
05:56The asset library generates our images, and it basically just follows the exact same flow described to the AI. For this next diagram, we're gonna go with the sketchy hand drawn xCalidraw look.
06:07I think we'll go with some blues for the accent. And for this one, I think we'll just keep it high level.
06:13If I open this up in xCalidraw, you can see now we have a diagram explaining how to install skills into Claude. So we get the skill file, we go into settings, capabilities, prerequisites, so this is just one time, and it explains all of the steps.
06:27Customize skills, click upload a skill, select the zip file, toggle it on, and then you can start a new chat, and it'll auto load it every time you want to invoke that skill.
06:39So just a simple diagram, but it's pretty cool to show you essentially what this skill can do. It can build out these types of diagrams. So if you need to teach something or build out some kind of flowchart or a proof of concept for something, this is super useful.
06:53If you just want to visualize an idea or a workflow for something, this can be really useful. And instead of having to sit there and create all of these boxes and type out all the text yourself, now you can just use this skill to create these diagrams really really quickly. And considering that you can use XcalaDraw completely for free, it's definitely something worth knowing how to use if you ask me.
07:13That brings us to skill number three, which is where we start getting into some of the more decision making and thinking related skills. So if I open up a new chat and we invoke the, uh, expand and contract skill, this helps you take a simple idea, make it as complex as possible, and then narrows it back down so that you can basically decide the scope of the idea that you have.
07:35So I could start with a simple idea, something like, I wanna build a fitness app that allows users to track their caloric intake, build out a meal and workout plan, and do it all using AI.
07:47And from here we're going to go through a series of questioning where it takes that initial idea, expands it as much as possible, and then contracts it back so we actually have a narrow scope and we can basically decide exactly what should be included and what should be excluded from the original idea.
08:04Now the first thing it's gonna do is create a list of around 25 different ideas that it thinks you could have, might be nice to have, you could consider later or completely remove. So the idea here is for you to sort all of these ideas into specific categories.
08:20So we could always revise this list if we want to, otherwise I'm gonna say it looks good and let's start the sorting. This first list of ideas are all of the ideas that we could add, but that are kind of just extra.
08:32So it's helping us brainstorm all of the possible directions we can go, so that we can eliminate all of them and keep the ones that are actually gonna matter. And don't worry, you don't have to sit here and reorganize the entire list. It's going to prompt you and ask you questions in batches, so that you can actually narrow this down.
08:47So the first question is which ideas are absolutely core that the app must have? So in this case, we'll say personalized meal plans, personalized workout plans, maybe a coach, and food logging.
08:59Honestly, probably all of them. Which one of these are core? Do we have any more?
09:04Maybe we'll do the barcode scanner, and we'll just hit continue. And I'm just gonna select a couple more to have in the core list.
09:11These again are the elements that are absolutely essential to having our app run. And then we're here for the next round of questions. So are any of these core, tracking adjacent features?
09:20Let's go with body progress, progress photos, auto generated grocery lists. We'll add a few more just for the sake of example.
09:29There's a lot of questions in this skill, but if you have something where you wanna flesh out exactly what should be in and what should be out related to your idea. Basically helping you decide the scope. This is one of the best ways that you can do it.
09:41It's gonna help you anticipate things that you didn't know you wanted, and it'll also help exclude things that you know you don't want before you begin working on something. Now we're working on explicitly out.
09:52Once again, I'm just going to select a few things that will explicitly leave out of this first version of our idea, just so that we can see how this works. Now we're on the maybe later round, so I'll just select a few. And obviously, if you're doing this for real, then you're gonna wanna answer the questions a little bit more serious than I'm answering them right here.
10:10When you add maybe laters, they come with a trigger. So this is how you decide when you might want to add those specific elements back into your plan. So in this case, if we're building out an app, we might want to add the manual food search with a database after we have a thousand paid subscribers for example.
10:27So you can add triggers for each one of the maybe later items that you add into your plan. So when we add adaptive plans that reshape weekly, maybe, you know, after three months of version one for example. When would we add the Apple Watch standalone app after our iOS app hits 5,000 users for example.
10:47Basically, you're trying to plan something out and you have a list of things that you might wanna add, this helps you figure out when you could add those down the line. This is a question heavy skills. There's a lot to answer, but it really helps you define exactly what the scope is going to look like.
10:59And so the output when you're done is going to look something like this. You have the core features that must be shipped with your idea. So in this case, if we're building out this fitness app, all of the things that must be included for this idea to work.
11:12We have nice to haves, maybe laters with triggers, so we know when to revisit some of those ideas if we still wanna add them into our product or whatever our idea is later on. And then we have everything that's explicitly not a part of the idea, but it's still nice to see some of those ideas or see the options and then explicitly say no to them, so that we know exactly what's in and what's out.
11:35It creates these sort of concentric circles. You can see what items land in the center, what are nice to have, maybe later's, and also all of the items that are explicitly out of this specific project or idea.
11:47And we can also stack skills here. So we could say, build out a better infographic representing the scope using our infographic generator.
11:55And here you can see we visualized our idea a lot better if we were serious about building this fitness app. Here are some of the things that we need. Nice to have course ins and outs.
12:04Now you can see that we're starting to stack some of these skills together. So if you're actually following and utilizing some of these skills, they really work nicely depending on what you're trying to do. If you're a visual person like I am and you want to see some of your ideas, this is really nice to have.
12:18Now one disclaimer that I want to make is that I didn't create all of these skills completely from scratch. A lot of them were inspired on someone else's skill, so I'll try and leave credit in the description next to all of the skills so that you can find the original creator of that skill even though these are my own versions, I still wanna give credit where it's due.
12:36That brings us to the next skill we have on the list which is the steel man skill. This skill makes sure that an assumption or an idea you have will still stand up even against the best arguments that AI can come up with. Ideally, you start with a statement here and then see how it plays out against the best steel man argument.
12:53So for example, we could say, I'm going to launch a fitness app that helps people create workout plans, track their meals, and it does it all using AI. And I'm gonna be able to scale it to $50,000 per month of recurring revenue.
13:07Uh, I'll vibe code it myself, and I'll be able to market it through paid advertising because it only takes around 2,500 members at a $25 membership to hit that goal.
13:18This is really cool because it sort of plays devil's advocate on both sides of the argument. Now if your statement's pretty vague, then it's going to ask you a bunch of questions to clarify your position. But since I gave it a pretty concrete statement here, it didn't need to ask me anything.
13:31And from here, it basically goes through and it makes the case for, the case against, and whether or not I'm winning that argument. Depending on how specific or vague your statement is, it might ask you some questions, but in this case since I was really specific, it was able to create cases for and against my idea.
13:50You can see how it sort of breaks down all of the arguments, so you can decide if you're thinking is actually right here. In this case the first angle is the market and the wedge. So how you're entering the market.
14:01And the case for it is that AI genuinely unblocks personalization in fitness and nutrition and it's not just vaporware. The subscription model is proven. The category is large but the case against is much stronger.
14:14So it's brutally saturated. There's MyFitnessPal, MacroFactor, FitBot and a bunch of other tools in the space.
14:20There's competing hardware and there's no ICP or wedge stated. AI fitness app is a feature not just positioning so I'm losing this argument.
14:29There's a few other arguments here. So the only 2,500 members math, 2,500 paying subs is genuinely relatively small relative to the overall market.
14:39So it's the case that I was trying to make. But the case against it is that fitness app churn is the worst in business to customer software as services. So it's around 60 to 80% drop off in ninety days.
14:53That means that I'm losing this case as well. It's gonna be very hard to actually get to 2,500 members. There's a bunch of different angles that it thinks about this problem from, makes the case for and the case against.
15:04And in this case, it looks like I have a pretty bad idea. But at the very bottom, it's going to kind of create a summary of all of the angles. Basically, the initial statement does not hold up and the idea needs substantial revision.
15:18If I was actually thinking about doing this, I definitely need to rethink about some of the positioning and the points that I've made because my idea needs a lot of revision. So it breaks down all of its points down here, but for me the main value is just seeing how it creates a case for, and then directly makes a case against each angle that it lists.
15:38For this specific example, I'm just not winning this idea, but hopefully you can start to see how this works. This next one's really simple and they use it all the time. I call it the Promptimizer and it helps you write a better prompt no matter what you're trying to do.
15:50So I could say something like, I need a prompt to help me draft cold outreach emails to potential investors for my SaaS startup. I hit enter, and it's going to promptimize my idea. Once again, depending on how specific or vague you are, if you didn't give it a ton of information, then it's going to ask you a few questions before it just spits out a prompt.
16:10And a few seconds later, we get a full prompt that we can use in a new chat session to get whatever result we're trying to get. We could do it for something more visual. So write me a midjourney prompt for a moody cyberpunk street scene at night.
16:23We can bang out the few questions and now we get a full prompt we can use. So I'll copy this and we'll come back into midjourney. I'll just paste it in and hit submit.
16:31And now you can see we get much higher quality results than if I tried to just write a prompt for something like this on my own. You can use it to get better prompts quicker and that's why I use it all the time. Now the next few skills actually require you to run them from within Claude code and although that sounds a little bit intimidating, it's really nothing too complex.
16:49In the sidebar, you just need to come over and select code. You're gonna select the folder that you wanna work in and that's also the folder where you'll want to install these skill files. For these, if you need to install them, can just drag them in and say help me install these skills.
17:02It'll read the skill files and then figure out how to actually install them so you can use them in your environment. You might need to restart Claude in order to see the skills pop up, but the first one I wanna show you is called Swarm Consensus. Now what this does is it allows you to present a question in the chatbot, and it's going to go out and ask all of the top AI models such as Grok, Gemini, ChatGPT, all from within one call and then summarize the output for you.
17:28So I'll show you exactly how it works, but it uses this tool called OpenRouter. This tool allows you to set up one API key and use that key to access almost any model. So you have access to all the frontier models like Claude Opus, ChatuchuBT, Gemini, and Grok, and a bunch of other models as well.
17:46Now it sounds really technical at first, especially if you've never set this up, but trust me, it's a lot easier than it sounds, and it's definitely worth it. So what you'll need to do is you'll need to create a workspace, and then in that workspace, you'll wanna create a new API key.
18:01I've got one right here which I'll have to blur out, but I'll also make a new one so that you can see how this works. You can also set a credit limit. So this is the maximum amount that you'll actually end up spending through credits.
18:12In this case, I'm gonna set it to $8 and I'll hit create. I'm gonna copy this, and then I'm gonna come back into cloud and say, help me add the API key so that I can use the swarm consensus skill. And then I'm going to paste in that API key.
18:25The idea for this is after we set the skill up, we can use the swarm consensus skill to get consensus of all of the top AI models on a specific question that we might have. Now I already have a key, so I'm gonna add this one as a new one. And it's telling me that I shouldn't have pasted the key in the chat, but it's fine because I'm gonna delete it right after I finish recording this.
18:46Now you are gonna be charged depending on the calls you make and the models that you're using. So the more intelligent the model, like if you're calling Claude Opus, it's gonna charge you more than if you're using some of the quicker and smaller models like Claude Haiku or Sonnet. And the same thing basically applies to every AI company whether that's Gemini or ChatGPT.
19:07The larger and the smarter the model, the more it's going to cost. Even But so, I've done around 30 sessions with the most intelligent models, and it's only cost me $15. Once the key is added, we can use the swarm consensus, and so let's give it a try.
19:19So we'll say something like, give me five unconventional business ideas for a one person operator targeting the 100,000 to $300,000 per year range.
19:29This skill is also designed to help you decide which models to choose. So in this case, it's telling me that since this is a creative brainstorming request, it's perfect for a cheaper tier with more models. In this case, we can run eight cheap models for just a few cents.
19:45We could also do five cheap models to get less noise or we can go with the frontier models if we wanna spend around 15¢ and we can wait a few extra seconds. But for this specific example, we'll just go with the recommendation and we'll see what it comes back to me with.
19:59And this is super cool because now it's running our original idea of giving me five unconventional ideas through eight different models.
20:06So we should get 40 different responses, and then it's going to synthesize all of those and tell me if there's any overlap and which ideas are the best. And now you can see the full consensus right here.
20:17So it's gonna give us the strongest most unconventional ideas. We've got the top five ideas right here, which honestly, dead brand revival, these are pretty, uh, unconventional if you ask me, and it also kinda helps us get an idea of what the different models were thinking.
20:32So where they disagreed and unique angles that were also worth thinking about. And to get this answer, it only cost 3¢. So super cheap if you ask me.
20:41Another example of how you could use this is for something a little bit more complex. So we could say, I'm a US citizen working remotely from Portugal for the next six months. So what are my tax obligations going to be for each country?
20:54If you have legal questions or something that's a little bit higher stakes, it's going to suggest you use the more intelligent frontier models. So in this case, I'll just go with the recommendation. And I'm also getting a follow-up question to provide more context.
21:05So I'll tell it I'm a self employed employee, I guess. And the frontier models take a little bit longer, but now we get a full breakdown with a consensus answer. So we have a much more detailed answer to our original question than we would have got if we had just asked one of these models at a time, and most importantly to me at the very bottom you also get where the models disagree because sometimes specific models will call out different elements that other models didn't pick up on, and that's what you wanna see over here.
21:32It's helpful to get four different answers on the same thing, and then see what the overall picture looks like, is exactly what the skill does. This type of consensus agreement is really helpful in a lot of areas for things like code reviews. You can also stack it on top of something, we could use it alongside the steel man skill.
21:49And we could see if four to eight different AI models all agree that these positions are actually correct. It's a slightly more technical way of thinking, but if you're serious about getting actual answers, you wanna do code review or idea synthesis.
22:02I've mostly been using it for creative thinking and legal questions, but if you wanna see that a lot of models agree or disagree on something, this is a great skill to use. The mental model is if you want a lot of different brains to agree on the same thing and figure out what that is, then you use the swarm consensus skill.
22:19But that brings us to the last skill which is the find skills skill. Now when you install this, it basically tells your AI agent to look on this website which is skills.sh. I'll leave a link to it right below, which is a database of over 90,000 skills that people have already made and built.
22:37This is a really cool website where you can come and scroll through the skills that people have already developed and published. And so chances are if you're trying to do something, there's likely already a skill that someone out there has created. And so it can at least be a good starting point so you don't have to build every single skill that you want from complete scratch.
22:56Like I said at the beginning, in a world where everyone has access to the same AI models, the skills that you use and the prompts that you use are going to differentiate the results that you get from someone who spent more time setting up a good system. One last note I'll leave with you is in the instructions for Claude, I've added this one sentence, which is improve and suggest.
23:16So while working note opportunities for automation, improvement, repeatability, and if a task is a good candidate for a Claude skill, tell me so that I can turn it into a skill and reuse that workflow later.
23:29This way Claude will help you learn when to turn a workflow into a skill so that you can reuse it later. But beside that, I hope that you found this helpful and as interesting as I did. I've actually installed a lot more skills that I didn't quite have time to discuss today.
23:42So let me know in the comments section. Did you find interesting? Was it helpful?
23:46Will you use any of these skills? And most importantly, let me know if you wanna see another video breaking down even more skills or a different type of AI automation type video in the future because I have a bunch of fun making these videos and I hope that you guys like watching them.
24:02And make sure if you want any of these to grab them right below that .
The Hook
The bait, then the rug-pull.
Alek opens not with his face, but with the output: a live interactive infographic already on screen. Then the stat: sixty-five percent of people are visual learners. The title has already done the credibility work — not "here are some skills," but "I tested a hundred and these seven survived."
Named ideas worth stealing.
01:00concept
Skills as Differentiator
In a world where everyone has access to the same AI models, the skill layer — configured workflows on top of the model — is what produces different results.
Steal forMCN+ positioning, JoeFlow marketing, any AI tool pitch
07:13model
Expand & Contract Scope Loop
- Core (must ship)
- Nice to Have
- Maybe Later + triggers
- Explicitly Out
Forces scope discipline by first expanding an idea maximally, then systematically eliminating and categorizing features. Trigger conditions on 'maybe later' items prevent scope creep later.
Steal forAny product scoping session, Mod Producer runsheet planning, MCN+ feature roadmap
12:35model
Steel Man Scoring
- Case For
- Case Against
- Winner verdict per angle
For each angle of an idea (market, math, channel, etc.), AI builds the strongest case for AND against, then declares a winner. Aggregates to overall verdict.
Steal forOffer validation, business idea stress-testing, MCN+ sales argument hardening
18:20model
Swarm Consensus Tier Selection
- 8 cheap models — creative/brainstorm (cents)
- 5 cheap models — less noise
- Frontier models — legal/high-stakes (~15c)
Model tier is matched to query stakes automatically. Creative = cheap + many. High-stakes = expensive + accurate.
Steal forJoeFlow /bakeoff skill, any multi-model orchestration design
How they asked for the click.
23:03subscribe
“make sure if you want any of these to grab them right below that big shiny subscribe button”
Soft and brief. Drive link in description. Subscribe CTA is framed as access, not ask.
Visual structure at a glance.
