AI Agents Full Course 59 Minutes (for beginners)

I think AI is confusing. There, I said it. I think there's a lot of terms, skills, MCPs, agent harnesses that are difficult concepts to understand. So I had my friend Remi come on the podcast and explain it in the most simple terms possible. In this free course on how to master AI agents, he breaks down exactly which each piece is, how they connect together and the simplest ways beginners could start using them today. Enjoy the episode. I begged them to come on Remi Gascals on the pod. You've structured your company where you basically have these folders and dot MD files that run your company. And what I want to do today is I want you to teach people in a beginner-friendly fashion. This is only for beginners how they could do the same thing, how they can set up their own executive assistant, head of marketing, chief financial officer. Basically, I want you to tell us the concepts behind all this. By the end of this episode, Remi, do you think you can do that? 100% Greg, we're going to go through all the concepts that make up an AI agent. By the end of this video, you will know exactly how you can build up agents to run complete departments of your life and your company within any agent platform you choose, whether it's Cloud Code, CodeX, OpenClaw, Manus, all of them. All right, let's do it. Sweet. So one of the reasons why I really wanted to make this episode is because I feel like the AI landscape is moving into stage two from chat to agents. And most people are getting left behind right now, just using the chat models. And the founders and employees that are utilizing agents know word of a lie 10 to 20 times more productive in their day. And when you stack that up over days, weeks, years, you're going to just be miles ahead of the competition. So I really want to make this episode today to help bring everyone up to where the AI landscape is at the moment, and to start using agents to manage every department of your business. So the key thing to understand here is chat models versus agents because the word agent is thrown around lots online. I'm sure you've seen it Greg like, AI agents this, agents this, use this agent for this, and it's kind of lost a lot of meaning. So I wanted to give start by giving a really clear definition of what an agent actually is. So the way I think of it is a chat model is question to answer, but then an agent is goal to result. So moving from just like you asking AI replies, then you do the work to you giving the agent a task, it planning out the task and then executing and then delivering you a result. Does that make sense? Crystal clear. I mean, the way I think about it is chat is kind of like ping pong back and forth, back and forth. And agent is, you know, you're giving it, it's a goal. I mean, the best way you're giving it a goal and you're hoping that over time it gets better and closer to that goal. Exactly. Yeah, that's exactly it. And I just think that's a nice way to lay down your head. Is chat is question to answer agent is goal to result. So when you chat to an agent, you might give it a task like build me a website for XYZ and then it goes away, it does its work and outputs this wonderful website to you. But it's really important to understand what's actually happening in this step here. So inside this agent step, we have what's called the agent loop. So you give it your prompt or task and it goes through these three steps here, which is observe, think and act. So let's just say, for example, we're actually going to do this demo after this, but if we gave the agent a simple task like build me a minimalist portfolio site for Greg Eisenberg, it's going to start by like you've loaded in that prompt. It's going to check if there's any files in the workspace that it can work with like maybe you've got some information on Greg Eisenberg. And then it's going to think about what to do next. It's going to act and then it just keeps going through this loop. So for that actual example of building the portfolio site for Greg Eisenberg, let's just say it was a blank agent. We hadn't given it any context. The first thing is it's received this prompt to build the website. And the first thing it's going to be thinking about is, okay, well, I need to build this website about Greg. Who the hell is Greg Eisenberg? So it's going to then decide to do some research into Greg Eisenberg. It's going to research everything about Greg and then feed it back into this observed step. So then it's going to think to itself, okay, so I've got this prompt to build a website. I've now got my research here, so I know exactly who Greg Eisenberg is. And then it's going to start thinking, what is the next step? And the next step is probably to write up a plan to build the website. So it might write up that plan, feed that back in. Now it's got the research, the prompt, the plan, and it will think, all right, what next? I should probably write the code, write the code, feed it back in. And it just keeps going through this loop as many times as it needs until it can conclude that the task is complete. And how it concludes that the task is complete is based on the parameters that you set in your prompt. So if you're giving it a research task, you might say compile 10 sources and then create a report as a PowerPoint. And then once it's compiled 10 sources and build the report as a PowerPoint, it can conclude that the task is complete and then give you the output as the user. The the agent itself is made up of these four components. So it's the LLM, which is the brain behind it. So think like, you know, Claude Opus 4.6 or GPT 5.4 or Gemini 3, it's the model. It's got the loop, which means it just keeps going until the task is done and doesn't stop after one response. So you're going from ping pong to like it continuing to go rather than you having to sit there babysitting it. It connects in all your tools and then it connects in all the context. And a platform that facilitates this process and basically facilitates this loop to happen is known as an agent harness. And all of the popular AI agent platforms on the market that you'd be familiar with are just agent harnesses. They're just applications where this loop is facilitated. And I want to actually run this little prompt I prepared earlier. I want to open up codex, Claude code and anti-gravity. And I'm going to show you this loop actually happening in action. So I've nicely prepared before the episode these three demo folders, which we're going to run in. So I'm going to open up demo one to work in the Claude code. And the way these folders work is if you've used, if you're familiar with any of the chat models like Claude and ChatGBT, there's a project feature, which is where if I open it up actually, trying to get dizzy with me switching tabs so much. But if we create a project here, it contains all your chats in one place. It allows you to upload all your sources here, which is your context. And then you can even add custom instructions, which tells it how to behave within this project. And that's also known as a system prompt, which we're going to dive into how to do this with agents as well later. But it's a similar concept that you'd be familiar with if you've used projects before. But instead of the project being here on the cloud, we're actually working within projects that are local and our computer. So I've just selected this demo one for now. Then we're going to run, build a minimalist portfolio site for Greg Eisenberg. And then this little bit here just tells it to actually spin it up like to publish it on the web in a preview mode so we can see what it's done. So I'm going to run that. So this is Claude code. Yes, right now we're in Claude code. And this is just accessing it through the desktop app for Claude. So I'm just going to run that. And then I'm also going to give the same prompt to codex here. So this is the codex app. And you can see same concept. It says let's build. We can choose a folder on our computer to work in like demo too. And then we're going to give that a prompt as well. And we're going to tell it to host it on a different one. And then also in anti-gravity. So you can see same concept going in selecting a folder. And then we will give it the prompt as well. How should people think about security and these different products? I like to think of security as in just like scoping what they have access to. So by default anti-gravity Claude code and codex they're very very secure because they're built by these massive companies that have a lot on the line to protect. And I just you know if you're building out these agents to manage different elements of your business like the other way I built one that does manages meta-ads. And obviously that's quite a risky thing to give an agent control over managing ad budgets. So it just comes down to like what you feel comfortable giving the agent. And also you can control what privileges or control what tool permissions as access to so that if it was compromised for whatever reason the worst case like isn't that bad. And that means you know just giving it like read only access to certain important platforms and stuff like that. Does that make sense? Yeah. Totally. I mean comparing it to like open claw which is like yeah. Which I want to touch on at the end as well because that's same thing just another harness but it's just like the Wild West. Cool. And one thing like a nice little analogy to think about these harnesses is what we're going to learn today is we're going to learn to drive. So we're going to learn about how to you know steer the car like how the pedals the brakes work the accelerator works the hand brake. But then once you know how to drive you can kind of jump in any car whether it's like a old Toyota a Range Rover and you inherently sort of know what to do. And that just comes down to understanding all these key concepts that we're going to go through today. And you can think of the agent harnesses like different cars and some of them will have better features like seat warmers and cruise control but it's all once you know how to drive you can pretty much jump in any of them and use them. So we've just got our thing over here building tar building the website for Greg and it's going through this agent loop right now. So you can see here it's actually decided that it's going to launch an agent to go and research Greg Eisenberg. And I've connected it up to complexity so it's now using complexity to research Greg. So it's going through its first step of the loop. And I imagine that codex has also done something similar here. You can see it's still working but it's gone. And started to build this out through the loop. I think Claude code does the best job of actually displaying that loop and allowing you to see what it's thought about compared to the anti-gravity in codex but it's all just going through the same sort of loop process that I described earlier. And I think when you said that when you say you hooked it up to complexity didn't it's not like you asked it to hook it up right it just sort of did it? Yeah because I've I've given Claude code complexity as a tool via MCD which we're going to get into very very shortly all about MCP which is just connecting tools up. So we can see that in anti-gravity it's gone you can see this thinking process it's gone. I'm now examining the current directory to figure out if there's an existing project or if I build one from scratch it's then going I'm now going to start to build this thing and then it's built the the website and it's given us a little local host preview here. So it's created this nice little portfolio site for for you Greg. What's interesting is like it's super minimalist and I mean it did its job right like it this is I would totally launch something like this. It actually looks really nice. Did it scrape your email address correct? That's that's not my email Jess and I don't want to talk anymore so but yeah so there's a few copy things but other than that yeah it's done a pretty good job but it did and if we go it's those anti-gravity if we go into codex as well you can see here it's finished doing its website which is somewhat similar I think I prefer Gemini's yeah I agree and if we check out Claude as well it's still going you can see this loop right it's gone okay it's first off who is Greg Eisenberg it's gone in research Greg then fed it back into that observed step and it's gone all right what next now I need to create the HTML file so it's written the the code and then now it's gone okay so he wanted it spun up on this local server so now I'm going to spin up on the server and then the last iteration of the loop is to check that it's actually done and can conclude the task is complete it's opening it up and screenshotting the website and then reviewing the screenshots to check that the website is complete and you see here it's done another pretty good job this one's very similar to the Gemini why it's true but yeah that's just like demoing how that loop is actually working in real time yeah I mean what comes to mind just by watching this is like how many people on the planet would benefit from a very clean website and like how do you set up these agents so that like you know maybe it's like a cold email loop right like you're sending cold emails hey I built you this website so into a business do you want it it's going to cost $250 yeah that's actually a great idea pre making websites for companies and it's like an off the shelf thing it's like hey I made you this website if you want it like if you want to own it's $250 you can just do a mass cold email thing cool so I think that's like pretty much illustrated that agent loop example so I'm just going to go back to our trusty board over here but you can understand that it's just like all these apps are just different flavors of the same thing and then what we're going to be working up to today is my workspace look something like this is I have you know a big like a folder of each company or client that I'm working in and then I'll have folders underneath with all my heads of departments and then within those heads of departments I have skills and MCPs which we'll get into and context and then I've got like an overarching one at the top to just to sort of manage them all but we're going to be focusing today on building out this executive assistant to take care of just your manual day-to-day tasks and free up at least once two hours extra per day cool so to build this out like we did with our demos it's running off your local files so we're going to create a folder here called executive assistant and also through building out this assistant it's going to allow us to clearly explain each of the concepts of building an agent in real time and the way I like to think about building agents is onboarding them like a real employee so if you took on a real executive assistant you couldn't expect them to come into the office and you'd give them a task without explaining your business first your clients what you do the tools because they just would not be a very good executive assistant so that's the first step that we need to go through when we're building out this agent so I'm actually going to work in co-work at the beginning so co-work is just another agent harness to do the pretty much the same thing as all the others just that loop connecting in your tools and the context so you can see here this was my little previous session where I was building some diagrams but we can go and you can follow along include code or codex or anti-gravity or whatever agent harness that you want to work in but I just think that co-work has really nice simple UI for people to just understand really well what's actually going on so we're going to open up this executive assistant folder and you can see here that if we ask it write me a cold email and send that off so people are going to ask how how did you transcribe you did like a voice to text yeah so that is um I use one called monologue but there's a lot out there on the market whisper flows another popular one and it just allows you to hold a little button on your computer and just yap away and it will just transcribe it neatly into text and I find it looks good yeah my blog looks really good yeah I think it's built by the team at every everything yeah um it's a cool product but uh so what it's asking so it's straight away we it's got no context here so it's working out of uh that folder here on our computer but there's nothing in the folder and it has no memory of our previous sessions um and it's asking like what like what do you even sell um and then we've got to kind of give it like who do you target what tone do you want this is all things that our executive assistant should know um so I'm just going to stop the response there and one thing that's really important to know which might be a bit of a shock moving from chat to agents is that these agents memory work a little bit different so if you're used to using chat models like chat gpt and Claude if you open up a fresh session in the one of these chats you don't give it any context you don't upload any files and you just say who am I and what do I do it's going to know a scary amount about you and that's because with these chat models they have memory built in automatically so every time you sort of say things that are important the chat model saves it to its memory in the cloud that you can't see and you can't control and with agents you have to set up memory and control exactly what you give it and I think that's actually a benefit not a limitation because what happens is if you're using chat gpt and it's got the auto memory you're having conversations about three different companies maybe you're asking for relationship advice and then all of a sudden um when you ask it to write a landing page copy it's pulling in context from all these other places that you don't really want in there so with these agents um you need to actually set up that context and memory so as you can see when we asked it to write a cold email it just had no idea about anything so we need to give it a context file and the way you do this right so you can see this example here it doesn't know anything about us and that's because we haven't populated what's called an agents.md file and an agents.md file is just like a system prompt just like if you've created any custom gpt's before you have that field for custom instructions or in the project like I just showed before you have that field for custom instructions and it just gives it this context that's kind of always there always on and you put in there things like its role context about you your preferences for working and then what happens is every new session before it answers your query or task it loads in all this context to its brain as part of that observed step in the loop so I have pre-prepared part of me so I've pre-prepared a agents.md file here so if we drag this in over here this when you're working within Claude code it's called a Claude.md when you're working within Gemini it's called a Gemini.md but when you're working within codex or open Claude it's an agents.md but it's all the same concept so we can drag this into our folder here and if we open up this file for a little preview we can see here I've got in here all about me what my business does my working preferences like the tools that I use and what for like notion project management stripe we've got you know all the information my item customized loaded with context here and I pre-prepared this but and if you want to make one of those you can just use Claude chat or co-work whatever and you can ask it to help you build out this agents.md file and so just ask you interview style questions to extract all the context from you and then build the file so if I jump back in now if I go to a new task same folder and we say write me a cold email it's going to have all that context yeah that's what we hope that's what we hope there we go and it knows you can see this files over here it knows automatically to load in this file if you tie look correctly yeah it's basically just like a reminder file yeah pretty much it's just like loading it in so it has all this set context before you even start working and one of the other big shifts to make which comes with moving from chat to agents is prompt engineering used to be the big thing it was like here's the ultimate prompt for going viral on social media or use this prompt for this and now it's all about context engineering it's about how well can you load up your agent with all the information about your business so that your prompts can be stupidly simple like write me a cold email and you're still going to get an amazing result you can see already here it's already asking like is it a brand or sponsor potential partner or consulting client so it's already got that context book of call you know it's loaded in everything that we've given it from that agents.md file and then now we've got a pretty decent cold email there ready to go so that's basically agents.md files for you and you want to create one of those to onboard your agent with all the context it needs and if you have lots of context without getting into too many advanced concepts here sometimes what I will do is I will create like a folder called context load that in and in here it's got different files about me brand voice idle customer profile etc etc and then in order to keep this smaller I will then just say in this cloud.md file before answering any questions or before doing any tasks read my context folder to understand about myself and my business because by default if you just have this context file in here but no cloud.md it won't load all that into the session by default but if you tell it in this file that it always loads in so then check this file you can start to like string all your context together and a lot of people have done that with obsidian so they'll have like in their cloud.md file they'll tell it to go check their obsidian vault for their second brand to go and find context so that is agent.md files explain so that's how you actually when you're onboarding your agent like a executive assistant you can train it up on who you are and your business and then as you can see here you know I've got folders for all these different roles in my business and in the head of marketing that cloud.md file would look somewhat similar but in the top it would say instead like you are my head of marketing you speak like this these are your tasks these are your roles and then the second thing here is about memory and the self-improving loop so we've solved the problem now trying to get too dizzy with these switching tabs but we've solved the problem of our executive assistant not knowing anything about our business but now we have a new problem which is it doesn't really remember the intricate details or your preferences across sessions unless you're manually going and updating that cloud.md file so you can see here if we go my favorite color is lavender it'll probably say something like got it noted yeah that makes sense right because it's and it's it's adding where is that adding it what's not adding it that's the thing so we can tell it my favorite colors lavender and it's gone the users just shared use that's that thinking step it's like the users just shared this like no nothing needed good to know I'll keep that in mind but then if we go into a new session same folder and we go what is my favorite color mind my spelling it's going to say I've no idea what your favorite color is even though we just told it and that is an issue you know because if you're working you know in you've got like a head of sails or something and it keeps it signs off your emails wrong and you tell it you correct it and you say never sign off emails with cheers say warm regards and it will go okay got it noted but then the next day you start working and it does the same thing again it's like what like my agents broken but really it's not it's running off those context files in the back and unless you are manually updating it it won't know to save that preference so what I like to do is I like to add in something like this to my agents.nd file so this is just a little simple thing you can pause the video and copy it but I like to I'm just going to remove that context file for now that was just to illustrate that example of adding more but we're just working with this one file for now so I'm just going to open this up so I can edit it and I will quite often add something on the bottom like that little snippet and this basically just says actually you know what I might just add it at the top just so it's there top of mind for my agent because I think this is really important so you can see I've just added this in it just says read all files in context read memory.nd this is what you've learned over time and then when I correct you or you learn something new update the relevant section in memory.nd and it's just got a couple little things here and it just says keep memory.nd current when something changes update it in place and replace outdated info so we can do command s to save that and then I'm going to add another file here we can actually just duplicate this and this one I'm going to call memory.nd and then we can open up this one and I'm just going to remove all this context here except I'm just going to keep those sections. So memory.nd is basically I mean it just it's what it sounds like right it's basically like you want to you know if the goal is to build you know AI employees that do things for us they're going to need to need to remember our preferences right a good employee remembers preferences and learns over time and not get back compound so memory.nd is just a place that you can just make sure that over time it you know your whatever you're using co-work whatever it ends up getting compounded getting smarter so ultimately you might be trying things like co-work and and you're you're not getting good results and a big part of that is you don't have a cloud.nd in memory.nd sort of yeah exactly exactly and now the thing is some of these agent harnesses have started to add in this memory system that we're doing manually telling it to update some of them have got that built in automatically like open claw and I believe like madness and some of the others have that built in automatically but it's still important to understand because it's just doing the same thing under the hood except they've just set this up for you. So we've got this here now we've got our memory.nd our claw.nd and then now if we go back into co-work if we do a new session in that same folder and we say my favorite color is lavender better remember for the sake of the demo I hope that it does what it's told it's gonna remember it you can see here perfect it's gone good I'll remember that let me save it to memory and now you've got this big memory file that builds up over time and whether for example this is your executive assistant so it might be saving preferences like how to sign off emails or don't connect with clients on Slack I always want to keep client comms on email but if you're building out like a head of marketing it might be preferences about how you like your ads structured in Facebook manager if you're building if you've got a folder where you're working on a website or an app it might be things like don't use dark mode and then it will update so it'll never use dark mode again and there's just compound over time so as you start to build up these rules the amount of errors go down and there's just compounds and compounds over weeks and months. Remi have you seen some of these memory.nd files get so big that at a certain point it's just ineffective. Great question I personally haven't had that happened to me yet I haven't hit that threshold but a best practice for those cloud.md files is to keep it around like no more than 200 lines and yeah I could imagine if you started to build this up over years and years you'd eventually hit a point where all the rules are stepping on each other's toes and then you know you could probably go through and do a bit of like a manual clear but I haven't hit that threshold yet. Cool so people don't need to worry about cluttering their memory.md. I wouldn't worry too much I mean if it's saving like this like this silliest little things like the tiniest corrections you can maybe update that cloud.md to say only save like substantial corrections you know and then you can have a bit more control about what it's saving so that's probably yeah yeah what I would do there but once you've set this up now when you say something like quit writing so formally it's going to do the task then update it's agents.md or in this case cloud.md to keep time casual never formal and then now in any new sessions it's going to keep that preference over time which is pretty cool. So now we've got our executive assistant set up with memory and we've given him his role we now need to connect our tools because by default most of these agent harnesses they just have web search baked in but if you want to actually start linking it up to your tools like Gmail calendar and everything else which is where the real productivity gains are made you need to do so via what's called mcp and I actually got Greg I got this mcp explanation from when you had on is it Ross Mike. Yeah so this he did a great explanation and it just dropped into my head really nicely and it's basically that before mcp's your agent or your LLM in order to speak to tools it had to kind of learn their language because called speaks English notion speaks Spanish Gmail French your browser speaks Japanese and Slack speaks Chinese and it was capable of connecting to those tools but it required these extensive custom developments that took a long time but then anthropic actually created mcp's that right yeah it's right yeah anthropic built mcp to basically sit as this translator in between your tools so that cloud can still just speak English and your tools can just speak their languages and this mcp speaks every language and then just translates your calls from your agent to the tool and then from the tool back to your agent so just set a really easy standardize way to connect tools up and that's what we're going to be using to connect all of the tools to our executive assistant so if we go back into co-work here you can see that cloud make it really really easy to connect up your tools you can just go to connectors browse connectors and they've got like hundreds of all the like biggest apps that you probably use and you can just you know add them sign in pretty self explanatory but I believe codecs would be the exact same you know you can go skills or if we go settings they probably have like a and then like Manus is the same for example if you're into Manus we can see we can go and connect our tools very very similar process and then same with complexity computer you know you got your connectors and you can connect all your tools in here it's just all using that model context protocol mcp so I've already before the episode gone and connected all of the tools that I use most like gmail google calendar granola and ocean they're all set up already as mcp's and what I'm going to do now is I'm actually going to open up this executive assistant folder in Claude code to sort of demonstrate how these harnesses are all the same and they work off your local files and the real future proof AI stack is just having those markdown files on your computer and the reason why I like to work in markdown files is because it's just the easiest sort of format for your LLM if you're agent to actually digest and understand compared to if you were to give it your files as like a docs or PDF file so I like to use Claude code within visual studio code so you can see here I'm just going to it looks very similar to anti gravity I'm just going to open up our executive assistant folder here and the way that I see the future of this ongoing Greg is I think that everyone's going to have their what I call an AIOS like an operating system and this will just compound over time like you saw with adding the rules and getting less errors but with adding your tools and then skills which will get into which is basically just training AI on your processes and I think that everyone's going to have like an AI operating system they work in and everyone will just have personal agents and agents to manage each department of their company and people won't actually use these apps anymore like I've connected up Gmail Google Drive calendar granola for my meeting notes striped for payments notion for project management and I don't even enter these tools anymore I just sit in Claude code as one central place and an example here is I sent myself before the episode I sent myself an email from a fake prospect and I also entered in granola a fake meeting with this prospect so now I can say things like summarize my inbox from today I'm you know someone might ask like well how important is that really you know like is that such a high value task like our what are high value tasks that you're actually getting done here so or maybe or maybe you get a lot of emails you know well if you know emails is a big thing if you do get a lot of emails but just having like all those tools connected in one place and not having to switch and copy paste context so you'll see an example here right so we've got summarize my inbox from today which is like one of the most basic agent tasks ever but we can see this is one I sent earlier we've got this one email here like our call today excited after your call once next steps so I might just say here um okay great I review my meeting notes with Malthoshi from today and then draft up the email sending the proposal and creating the strike payment link and then go into notion and set up the project and where this starts to compound even more is when you start to build out skills for each of your processes because every time I do a process even like this manually prompting it um and I know I'm going to do it again at some point I'll then just turn that into a skill and then you eventually end up if you automate like three to five tiny manual processes each week with skills you eventually end up um automating like your entire life with these agents right so it's it's not so much in like summarize my my emails where it's super super valuable it's like that's where the starting point is and then we want to like manipulate it and use it and go deeper and stuff like that that's when exactly these tools really really are valuable and you can see here it's now connecting on my tools so it's going into Grenola and found the the full meeting of what we went through today it's now going into Stripe to create the product link um yep and then it's going into Notion to set up the project and then it will it should create the draft ready for us to go to send out that this is really like a new way of working right yeah it's it is it's so new and I and I even this task it's really simple right just sending an email based on a call with the proposal link and stuff but like even if you can just do something like seven times faster without having to go into all these tools copy the meeting notes into the page to give it context on your meeting it really starts to compound then you start to fit like a week in a day and then seven weeks in a week um and stack that up every year and you're going to be miles ahead of everyone else uh and when we get into skills you're going to see how this continues to get even better but um you can see here it's drafted the email it's pulled in all these insights from our call in granola which is like where I do my meeting notes and then it's created the Stripe payment link here ready to go that's cool and then now I can just go um send this email and we use my gmail integration to to go and send it uh and then that's really cool it is a I think this is the new way of working and Cody Schneider who had on the pod the other week I saw a tweet from him and he said that in the future everyone's going to have like an AI operating system like this and you're going to have like the 100x employee because everyone will come into their role with a pre-existing AI operating system and then build out skills for all their manual processes similar to how are describing and just keep building skills each week for anything manual that comes up until eventually their entire life and work life is automated great as you can see here it's now created the draft here in gmail ready for us to go and if we're happy with it in the platform we I could also just ask Claude to send it there and then but then what like also gets really cool is I'm going to demonstrate now how I actually build out skills for these processes so I know talked a lot about skills so far I want to just give a little overview on what skills actually are yep so the easiest way to think about skills is SOPs for AI so standing operator standing operated it oh my god standard operating procedures for AI so it means once you explain something once you never have to explain it ever again an example of this is without skills if you are creating a proposal for a client and you're sitting in your cloud chat or whatever agent harness you're using and you ask it to create this proposal you're probably going to go back and forth a bunch of times uh remove change the formatting here use this color blue for this part um put the price at the bottom instead of at the top and eventually maybe after 15 minutes half an hour you land at a proposal that you're really happy with and you send it and then next week you want another proposal written but unless you're going and finding the same session and working in that same session it's going to have completely forgotten all of these preferences and even if you have that memory system set up these kind of things you don't really want clogging up your memory they're better off as skills which is basically it packages up that process into a dot skill file and in that dot skill file it's basically just a marked down file that explains the exact process that you went through so you could create a proposal skill and then every time you now you need a proposal written it just takes that skill knows exactly what to do and then you can have that proposal the same way every single time so as a skill like a memory file like what's the difference between essentially a memory dot md and a skill like is it just is a is it just like a memory dot md file for particular job to be done that's pretty much like exactly it and all of these agent harnesses pretty much now have skills as a feature so you can see here like if we go into codex for example they've got skills here same with Claude as well and these when you're working with these agent harnesses that operate like mostly locally off your computer you can see here it actually operates out of this hidden file called a dot Claude folder skills and these are all of the skills that I've created there's tons and if we open one up for example like this one here let's find a good one for example I've got this one here for writing viral hooks and in this skill we have a dot skill file which is basically like your memory dot md which explains the exact process for writing viral hooks and then it's also got packaged in here some references like hook formulas okay so wait so how did you create that skill okay so there's two ways that I find useful to create skills is one you can have an idea of a skill you want to create off the bat so like viral hooks for example I had this course on viral hooks which I transcribed put it into Claude and Claude has this by default it has a skill creator skill added into it same with all of the a major agent harnesses they'll have a skill creator skill so it's kind of like skill-ception you use the skill creator skill and you say hey take this course on viral hooks and create a viral hook skill and it can create it like that that's one way and then it will package it up nicely with that skill dot md it'll do the whole thing for you wait a second you you asked you asked it to take the course yeah I uploaded the course like the full transcript upload literally and I said yeah based on this course on viral hooks build me a viral hook skill and then now I use that for my content team so just like we're building the executive assistant folder I've got a folder called content team and I've got that that uses that skill for me and the second way to create skills is going through a process manually once with Claude and then if you know you're going to have to do it again like that proposal example you can just say once you've done the task hey create a skill for what we just did and it will package up that process you went through and that's the second main way that you can create skills so in your viral hook example if you go into that folder again yeah so you have a references folder yeah so that is probably like was that yeah let's open it I'm just curious to see here it's got a full thing about like I was just from like four years bit screwed yeah this was basically from like a little course I put into it and did you ask it to create a reference folder like how should people think about no it just did it it just did it so I think what would be great is if we could actually demonstrate building a skill live with it and so for example like this process here I might I could create a skill called like a daily brief skill you know that goes through and summarizes like your calendar your inbox and your projects in motion and plans out your day for you in the morning and then you can run that on a scheduled task because a lot of these agent harnesses now are starting to introduce scheduled tasks so you can just run it on 9am every morning use my daily brief skill to prepare me for my day but I think another cool one here just to show you an example right of how my new like how intricate I make these skills is let's just say for this fictional meeting I had with this person I might say can you draft up an email I want to refer Moltochi to my good friend Sebastian who has an AI automation agency and can help them out better with their needs and then we can just go um Sebastian's email is and we can just say that right and then now it's going to be able to take the notes from granola all the context and then draft an email connecting these to a prospect with a friend and you know I have different like referral things set up like that with people with marketing agencies and that's just like a little manual process there tiny it maybe takes 15 minutes out of my day but then I can just go I want you to use your skill create a skill and create a Sebastian refer skill so that whenever I ask you to refer someone to Sebastian you know exactly what to do and you know his email address and then that'll build out that tiny skill for us tiny process but it means I like I know in the future I'm gonna have to refer someone to Seb again and even if this skill now saves me 15 minutes another five or six times they start to compound when you create skills for every single little process in your business yeah I guess it's like we should just be asking ourselves like you know in our day-to-day life like what are all the jobs to be done yeah what are all skills that we need like what are the repetitive repetitive processes or SOPs as you talked about and then setting up as many as possible right exactly exactly and just to give you a little demo here so I sort of alluded to that folder structure at the side of the video and this is it here so we've got work spaces I have with Remi and for example I can open up my content team and within this folder this is just like a little more elaborate version about executive but we've got our claw.md in here which explains you are like the main orchestrator you have these subagents it's just a more elaborate version of that claw.md but I've got a skill within this for like a meta ads analysis so that that was a process for example if you're a marketing agency owner this is probably like the kind of stuff that you can get inspiration from like ads analyzing you know taking competitors ads libraries breaking down all the creatives and their landing pages so I built out this ads analyst skill right literally just do ads analyst and then I paste in like the ads library URL like that and I'll click run and I didn't examine yesterday with the UDI which is a super large e-com brand and it ran through and basically scraped all of it took screenshots of all the landing pages it went and scraped all of the ads that are running all like 220 it then did a full deep dive here on all the ads visual analysis copy analysis why did this work what could be improved it basically did a breakdown of all the landing pages with screenshots and it did a master report here about everything that's going on so did just did a full breakdown and that was like a manual process that I would have gone through when I used to run like my marketing agency and that probably would have taken me like three or four hours and then I went through to build out this skill I went through the process once with Claude like I started a fresh session and I was like all right go to this ads library URL scraped this do this do this do this for like two hours and then after I've done the entire process I just said use your skill creator skill to make a skill for ads analyzing and package up the entire process we just went through as a skill and then now whenever I want to do that process again I can just invoke the skill and it knows what to do she's pretty cool crazy crazy absolutely crazy so now the the refer Sebastian skills live and now whenever I want to refer someone to Sebastian again I can just say yeah refer to Sebastian and it will just start to use that skill that's example their tiniest process but you build like those up for all these little tasks you do day-to-day and then it just compounds and compounds and compounds and I can already think of an idea here where you could then you can change skills together so you could example have like a meeting prep skill that prepares you for a meeting by researching the guest and compiling some talking points you might have like a podcast research skill for example Greg for a guest that's coming on and you might also created a morning brief skill and in the morning brief skill you can say if there's any meetings coming up or podcasts in my day use the podcast research skill to research the the the guest and you can like chain them together and build like some really really cool workflows yeah and you can have it so it send you an email right yeah exactly and then now these harnesses assigns get more and more autonomous like they're starting to add like you know in the car they're starting to add like cruise control and stuff like now within most of these harnesses you can schedule tasks like in co-work or cloud code now and you can like for example this one here you can go new task and I could say like run my morning briefing skill and then set that to go every every morning at 9am and then now it's like an automated workflow that like you've just got running every morning now which is pretty cool yeah I'm doing this right now like for example I'm buying a new car right now and it's like a particularly unique like color that I want and feature set and there's just none none really available so you know every three hours I I'm scraping all the different car marketplaces and then I'm getting a notification that you know when something comes up and it's it's crazy right like it saves me I'm one of those people that like if I didn't have this I would be spending an hour of my day just like checking religiously every single you know carmax and cars.com and auto trader and all these websites and refreshing like a insane person so yeah the schedule is great great example there but you know this like this is a skill that would be relevant for my executive same with that car one that could be a good executive assistant skill but then I've got those more elaborate skills built out for like you know my content team that adds library scraping one and then I've got like you know research weekly research skill for my newsletter team and that runs on a schedule every Thursday morning to go and scrape like I've built the skill out so it goes and scrapes Twitter and read it to find what's new in AI but yeah skills are so so powerful combine them with like your MCP so it can use your tools and then you can start to just train up your agent on all the processes in your business and I did a a build out on OpenClore for a agent to manage meta ads and it went pretty viral and the way I built this was with all these key concepts so OpenClore functions the exact same way so I I hope it hasn't timed out but I've remote accessed into my OpenClore dashboard here and you can see it's just operating off an agents.md file in the back end but instead of the .clore folder it's in the .openclore folder and then it's got a couple of these other ones here it's got a memory.md it's got some of these other ones that it's added on like a soul which tells its personality and an identity which tells it who it is but it's that same concept of markdown context files connecting your tools and then creating skills so that meta ads manager one that went pretty viral I just planned it out with Claude I was like I want to have this OpenClore manage my meta ads help me write the agents.md file to tell it you are my meta ads media buyer you do these processes and then I created skills so I created a add creative skill so you to go look in the drop box folder and create creatives I created a copywriting skill so I knew how to write good copy for the business and I just built out there's probably maybe 15 different skills and then I would combine schedule tasks which is cron jobs with skills and the context files and then just give all the tools it needed and just following the same process as we just went through to build the executive assistant I had an OpenClore meta ads media buyer which was sick. I love it and for the beginner are you like would you recommend people you know use OpenClore or should they be using co-work or man-ass and some of the ones you showed. So great question I would say that OpenClore is probably like one of the hardest to learn and set up of these harnesses. I would say co-work is probably the easiest I think complexity computer you did a video on it it's pretty simple too same with man-ass but I would definitely learn and get comfortable using like Claude code or one of these other ones before I started to play around with OpenClore and I would also have all the processes built out in Claude code first so for example that executive assistant over the next two weeks I might build out a bunch of skills like the Sebastian Refer skill like a daily brief meeting prep etc etc and then once I'm happy with how it's all functioning in Claude code then I could look to migrate that into OpenClore where it has that more autonomous nature to it so that's kind of how I think about using OpenClore and those other harnesses. Yeah cool all right anything else you want to cover. I mean like really there's no right or wrong way to run these like that was the executive assistant I've got one built out for all the other departments in my business and then other businesses I work on I have the same and you can just kind of build out that structure with what works for you and you've got like one other thing to mention is global versus project level which I'll just go over super quick so like those skills for example you can add them at a global level which means they apply to every single project you work in whether it's the executive assistant you're head of marketing and some skills you want globally because you might use them in every chat like a truncate skill that I created which just makes whenever I want to make something shorter it makes it shorter without compressing the sentences but just removing sentences that don't need to be there and that's something I want in every session so I've got that in global but you can have project level skills like that Sebastian Refer skill I would not want that with my marketing head of marketing because it's just like clogs up the context and you don't need it there so I would have that as a project level for example and you can have global skills versus project skills global cloud.md versus project cloud.md and same with MCPs you have global MCPs and project MCPs that's probably the other concept to go over but look other than that that's pretty much the entire agents crash course so it's just that loop running in the backend to complete your task and connecting in your tools your context and the LLM all in one place and I would just say to work out what roles you want to start to build out an agent for go into claw to your favorite chat model and get it to help you build out those context files through an interview style process just say ask me questions to build this out I'd connect all the tools that you need and then start building out the skills through daily use and then pretty soon you can have like pretty powerful agents built for every single aspect and department of your business. Remy thank you so much on good links in the show now it's in the description where you can go follow him get to know him a little bit better and I appreciate you coming on dropping some sauce thank you man thank you so much for having me on Greg it's been a blast.