YT YouTube2Text

YouTube2Text → Transcripts → The Official BMad-Method Masterclass (The Complete IDE Workflow)

The Official BMad-Method Masterclass (The Complete IDE Workflow) — Transcript

by BMad Code · 14,441 words · 796 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:00Hey everybody, my name is Brian and I am the creator of the BMAD method. This is going to
0:05be a long video, but I want to show everything involved with using Claude Code, all the agents,
0:10all the options, all within the IDE. This is the first time I've done this, and I want to show you
0:15so many details and just how cool some of these features are that I've never demonstrated before
0:20in a video. And I've never seen anybody else show some of these techniques also. But there
0:24are chapters here, so feel free to jump ahead to the chapter that you need. But honestly,
0:29I really suggest watching this through from beginning to end. There's so much good information
0:33that I'm going to be sharing with you today that this is this is just going to be so beneficial
0:39for you. You might not be using Claude Code, but there's going to be a ton of information in here.
0:44Trust me when I say if you can follow this, this applies to all of the other IDEs. So, install
0:50it for whatever IDE you want. A lot of what I'm going to be showing applies to all of them. Also,
0:56please give it a thumbs up. It helps the channel. Subscribe, tell your friends. Share this. Spread
1:03the word about this. So many people are using this in their day-to-day lives. Um, people tell
1:08me every day they're using this at work. They're using it for so many different things that I never
1:13even imagined. Let's dive right in. You always want to come to the BMAD method and see what's
1:18new. If you come to the BMAD method, this is also the easiest way to get started. And I really do
1:23suggest that you start out here. you can come to the read me and you can see here's the link to
1:28the YouTube where I'll have this video and other tutorial videos posted. Also, while you're here,
1:33if you're a fan of the BMAD method or if you try it out and enjoy it, I will say come up
1:37here to the top and then over to the right of the sponsorship button. There's a star over here.
1:41Please click star. This helps other people also find out about this project on GitHub when they're
1:46looking for agentic tools and helpers to help take their vibe coding to the next level. And this is
1:52going to be showing you how to do a green field development of a very simple to-do application.
1:57As always, I say take this method and make it your own. This is not meant to be the only way to do
2:02it, but it is meant to give you a framework so you can mold it to your own way of working, your
2:07own type of project. So, you'll want to subscribe to the channel, hit the notification bell so you
2:12know exactly when that video comes out. But first, let's get to it, everybody. All right. So, aside
2:17from uh these links here, if you scroll right down in the read me, you're going to see this workflow
2:23user guide. This user guide is very simple to follow. You can read this, but if you scroll down,
2:29look, there's two diagrams here, and they're very easy to follow. This shows you exactly how to use
2:33the BMED method, whether you're doing it in the web or in the IDE of your choice. First of all,
2:40you have a project idea in mind, right? You have two choices. You can use the analyst. The business
2:45analyst is here to help you further refine your idea before you go to the PM and start making your
2:52PRD. So, I'm going to show you how to install the VMAD method and actually get set up in cloud code
2:57to start using the analyst. And I'm going to show you the optional brainstorming and then also how
3:02we create a product brief which will then make the rest of the flow so much easier. And what's
3:07so exciting is you can actually do everything with the BMAD method now all within cloud code,
3:14all within the IDE. So if you so choose, you do not have to use the web and the full team
3:20stack anymore if you do not want to. I actually enjoy and have had great success using the BMED
3:26method fully in the IDE and I'm going to show you how to do that today. Really quick though,
3:31we have the analyst, right? And then it's going to go to the product manager which we will then
3:37go to the PO and then we will come down to the development phase. But take your time really
3:44understanding this. It might look complicated but it's actually a very simple flowchart. This is if
3:50you want to use the BMAD method for a brownfield or green field development to take it from
3:54ideation all the way to actually developing it in the code. Right now you see the only thing I have
4:00here is the BMAD method. Now, you do not have you do not have to install the full GitHub repo BMED
4:06method locally. You will only clone that if you want to create your own expansion packs or maybe
4:14make changes to the core system or contribute to the project, which I would love by the way,
4:18but you don't need to do that. I'm going to show you how to use the BMAD method in cloud code
4:23without even cloning the GitHub repo. So, first of all, let's um create a project. But what we can do
4:31is from any directory. I suggest doing it from the dev directory. We are going to type npx and then
4:37we're going to type bmad- method install. We are going to use the bad method 4.33.1. We will just
4:44hit enter to accept that. Let's enter the full path of our directory. Since we're in the dev
4:49folder, I'm just say dot meaning this folder. And then I'm going to say slash. And now I'm going to
4:54put the name of the new directory that I want to create my project in. Since we're doing this from
4:59scratch with no starter project, I'm just going to say simple to-do. That's the name of our project.
5:06Just a very simple to-do application. And hit enter. It's fine if that directory doesn't already
5:11exist. If you're doing this in a brownfield or if you already have started your project
5:15maybe with a template, you'll just give it the directory to your project and it will add this
5:21to an existing project. So then right here, this is a multi select, but we're just going to go with
5:26the default and select the BMAD core. The BMAD core is the agile workflow AIdriven that we're
5:34showing here. So just hit enter. And then we are going to shard the PRD, which means we're going
5:40to generate a large document, but then we're going to split it up into smaller pieces. So just accept
5:45the default here. And then the same thing with the architecture. The architecture file can be
5:50very large depending on what type of project you're building. And so you'll want to shard
5:55the architecture into smaller documents and this is all automated. I will show how this works. Um,
6:00but that is really going to help you maximize the context in your agents so they only load what they
6:07need instead of having to load and read through a whole massive file. They'll know which files
6:11they need in the moment to load and it will help them with their context. So we'll say yes to that.
6:17Now, here's a warning and this is just pointing out this is also a multis select. You can select
6:22one or multiple IDEs that you want to install to. For example, let's say that you are a fan of
6:28cursor and cloud code. Well, you can hit spacebar and cursor and spacebar and cla code. Today, I'm
6:35just going to be doing this in vanilla VS code. So, I am only going to be using cloud code. So, I
6:40have highlighted cloud code with the radio button. Make sure that you actually spacebar to hit the
6:45radio button. And if you just highlight it like I have right here right now and hit enter, this will
6:50actually not install it for any idees. Uh, one other quick tip is depending on the size of your
6:55screen, you might not see all of these options here. So, just know that you can arrow up and down
7:00and more options will appear and we're constantly adding more idees for support here. Hitting enter.
7:06And now this is asking if we want to include the pre-built web bundles. The web bud the web bundles
7:12are not needed for working in the IDE. This is kind of warning us in case we thought it existed
7:18and we typed in the wrong directory. For example, you know, we could choose a different directory
7:22location or we could cancel and start over. But in this case, we're going to say create the directory
7:27and continue. Okay. So, if we scroll up here, we can see what did it do here. It installed some
7:34commands for cloud code. So, these slash commands will be how we activate our different agents.
7:41And then we've also installed some commands and these are great because sometimes we want to do
7:46something without necessarily invoking an agent. We are now done. That is the install of the BMAD
7:52method. I know it took a few minutes to explain it because I was going through the options, but let's
7:57see how fast this literally is. I'm going to do it one more time. I'm going to do n px vad- method
8:07install. And I'm just going to give this a different folder this time. So I'm just going to
8:12say uh let's say dash demo one, right? Just demo one. Enter. Enter. Enter. Enter. Select my IDE.
8:22Enter. Enter. Enter. Boom. That's done. 5 seconds and you're ready to go with the BMED method. It's
8:29really that simple. All right. And just to see that our two projects are there. If we do ls, we
8:35can see that we do have the demo one, the one we just did, and also simple to-do. Simple to-do is
8:39the project we're going to use. So, I'm just going to do rm-rf demo one. We do not need that anymore.
8:46And now we can see that we have our simple to-do. So, I'm going to go into simple to-do
8:55and I am just going to open up uh VS Code. Now, right now we have no project here, right? All
9:01we have is what we've installed for our agents under BMAD and our custom tasks or commands for
9:10cloud code. And then we have the full BMAD core system here. Before we go on, I do want to say one
9:16other thing though, and that is I want to thank everybody that has actually taken the time to
9:21buy me a coffee. I've talked to a few of you and just know how much it really means for supporting
9:28me and the BMED method and the community here. So, I want to thank everybody that's come and
9:34supported the BMAD method. And if you would like to uh you can go to buy me a coffee,
9:39look for BMAD code or the link in the description. Any amount helps and I truly thank all of you. I
9:45like to use claude code in the terminal inside VS Code, but it doesn't matter if you use claude
9:51code outside. You could use it inside the warp terminal or your own external terminal,
9:57uh, you know, T-Max or anything, whatever you want to use, but I just find it convenient to kind of
10:02keep it all in here. It's easy to, you know, just have the, uh, keyboard shortcut to show and hide
10:07the terminal. Um, but we have no other project, right? So, as always, if you've used cloud code,
10:13you should know how to do this. You just want to type claude. And we are going to launch Claude
10:18Code. Of course, you will have to have installed Claude Code and set up your authentication,
10:23but I'm assuming you've done that. If not, go to the Cloud Code website, see how to install it,
10:29figure out whether you're going to do the $20 plan on the 100 or the 200, and then you're
10:33ready to go. Oh, by the way, I have been talking to a lot of people that have just been using the
10:38$20 version of Claude Code. If you're fine not having the sonnet, excuse me, the opus model,
10:45the $20 clog code will actually get you very far depending on how much you're doing. Uh,
10:50so just say yes, we do trust this folder. This is the first time we've opened this in a project. You
10:56could run the slashinit command, but I'm not going to do that yet cuz there's not really anything in
11:00this project for it to know about. After we build a little bit of our project and get some stuff in
11:06place, that's when I would do clait. Well, if we go back to the diagram, we can see that the
11:14first person or the first agent that we want to talk to is the analyst because we're going to do
11:19some brainstorming. And if you just hit slash, if you selected cloud code in the installation,
11:25you're going to see all these all these options here. And if you don't want to arrow through this
11:29list, what you can do is you can just start typing the part of the name of the agent that you want.
11:36So, for example, analyst, you can just start typing it like that and it will highlight it.
11:43Or if you know what it is, you can just type the whole thing or you can arrow up and down
11:48to the one that you're looking for. So, here's the analyst. So, we're going to select it. This is how
11:52simple it is to use clawed code with the BMED method. You saw it takes 5 seconds to install,
11:58maybe 10 seconds if you want to read everything. So, this is Mary, our business analyst. I know
12:03a lot of people think the power in the BMED method is in the development of actual code,
12:12but I think Mary the business analyst is probably the most special agent in the whole BBA method.
12:20If you are just trying to think of something to build or maybe you have a little bit of an idea,
12:26I really strongly suggest that you try using this business analyst with the brainstorming method.
12:33This goes beyond software development. You can use this for any type of brainstorming, whether it be
12:39in your day job or, you know, life questions. I've tried many different brainstorming tools,
12:46including professional brainstorming tools that have been developed by Google. For example, in the
12:51Gemini gems, there is a default brainstorming gem that you can use. It's it's nothing like this. So,
12:58you know, let's let's do that. So, we're talking to Mary right now. Mary is active. As always,
13:03you can do star help. And so, now these are all of the different things that are available to
13:07you. Now you can either select a number or you can type out the name of the command or you can
13:16rely on the fuzzy matching and usually the LLM will get it right. So I'm just going to select
13:22number five though which means brainstorm and I'm not going to give it a topic. I'm just going to
13:27say number five. Even though Cloud Code supports a context window that seems to allow you to go
13:33on and on and on without ever getting close to compaction, that doesn't mean you should. So,
13:40when we're done with the analyst brainstorming, it will give us a brainstorming document. It will
13:45save it to a folder called Docs, I will start a new chat. I like to start a new chat every time
13:53I have an output document, then I'm ready to go. So, even if I'm going to talk to the same agent,
13:58I will clear Claude or start a new Cloud session and then reload that agent and start going. So,
14:04here we go though. We're going with uh Mary and first she's going to ask us four questions. So,
14:09we are going to use our Super Whisper. What What is this brainstorming about? I would like to
14:14create a simple to-do app to demonstrate the BMAD method. So, this should be a very simple NodeJS
14:22background application. But what I really want to brainstorm about is some of the ideas we can
14:27do to at least make it a little bit interesting. So let's play some of your brainstorming games
14:32and see what we can come up with. So we'll let it spit all that out. So question two, are there any
14:36constraints or parameters that I should know about? Well, I already mentioned that we're
14:40using Node.js and I want this to be a very simple, straightforward small project just to demonstrate
14:45the BMAD method with Node.js and TypeScript and then we will submit that. So this is how you work
14:50with the agent. One of the magical things about the BBN method is it not it is not just like a
14:55taskmaster or some of these other things where you just throw out your idea and then it does all the
15:00thinking and working for you. As a matter of fact, what you're going to see with this brainstorming
15:04is it is actually going to push on you to get your ideas out. It is going to be your coach. So that
15:11is what is special about this. It is not checking your brain at the door and doing it all for you.
15:17It is going to work with you and it's going to bring the best out of both of you. The
15:22BMAD method, by the way, is all about elevating yourself, learning, and also elevating the LLM.
15:29So, as a whole, you are both better collectively than each one on its own. That is the special
15:35sauce of the BMED method. That's that's the secret everybody. Are we aiming for broad exploration of
15:41possibilities or more focused ideiation? Let's go broad. Let's go crazy. I want to show just what
15:46you are capable of as the amazing brainstorming agent that you are. Nothing wrong with flattering
15:52your LLM every once in a while. They're sure going to do it to you. Yes, definitely. And that is also
15:59the default. But when we're done here, this will take all of our brainstorming and give us
16:03this amazing artifact. This is one of the coolest artifacts in one of the most recent additions to
16:08the BMAD method. This was not even in the original release of V4. I thought this would be a V5 thing,
16:15but I love the brainstorming document output, and I think you'll love it, too. We haven't
16:20gotten to the brainstorming yet, but here we go. You have four options here again. So,
16:24you can pick the techniques, and there's a whole list of 20 brainstorming techniques built into the
16:28BMED method, or you can let the agent recommend it based on your project. Or you can do random
16:34creative chaos. This one is fun. You never know what it's going to get, but sometimes it picks
16:39some of the wackiest ones. or the progressive creative journey where we start super broad with
16:44wild what if type scenarios and there's five of those and then gradually go down to focus. I just
16:52want to show you all of the different options that we have here. So let's go to number one this time.
16:56All right, so here we go. You can see that there's quite a few here and there's actually more coming
17:01in the future. So we have what if scenarios, analogical thinking, reversal inversion,
17:06first principles thinking, structured frameworks such as the scamper method, the six thinking hats,
17:13mind mapping, collaborative techniques such as from the improv world, there's yesand, there's
17:18brainwriting, roundroin, random stimulation. There's deep exploration with the five W's,
17:24a classic brainstorming technique to really do a root cause analysis. There is morphological
17:30analysis, advanced techniques such as forced relationships, assumption reversal, role-
17:36playinging, timeshifting, resource constraints, metaphor mapping, questiontorming. But with the
17:41power of the BMAD agent, Mary here, she's going to guide you through these and help you come up with
17:46stuff that you never even probably imagined. You're going to see how with brainstorming,
17:51we can even make a to-do list super interesting and exciting. If we would have selected one of
17:56the other techniques up here, those also involve multiple selections from here and it will work
18:03through each of them one at a time. Usually 10 to 15 minutes is what it takes for each one.
18:09So this can be a lengthy process, but look, if you're at the beginning of embarking on a
18:14really large project or you have some ambitious ideas, it is worth doing this brainstorming to
18:19really come up with, you know, super creative just interesting ideas. take the time here to build for
18:26the future. Let's just pick a couple. Somebody in the community was suggesting that they would
18:31love to see six thinking hats and also the five W's built into the BMAD method. And then they
18:38realized that they were actually already in here. So just for them, I'm going to select number six,
18:45the five W's. So we'll do 11. Uh let's do role playing. Role playing's always fun. So now the
18:51agent will remember our choices and it's going to work with us one at a time. And remember here,
18:56we are not asking Mary to brainstorm and come up with this for us. She's going to guide us through
19:02this exercise. This is your own personal coach to brainstorm on complex, creative, or interesting
19:09ideas. Folks, this is worth I don't even know what. You know, you can pay experts to teach you
19:16how to brainstorm or be your brainstorming coach and it's going to cost you thousands of dollars
19:20and you're getting this for free and we're doing all of this in claude code. So, this is amazing.
19:25So, okay, what do we got here? Well, excellent choice. So, first we're going to start with the
19:29first selection number six, which was the six thinking hats, five ways, and then roleplaying.
19:35This is going to generate some seriously creative ideas. And again, this does not have to be about
19:40apps. You can brainstorming about everything. Maybe you want to brainstorm ideas for a vacation
19:45or brainstorm how you're going to retire in 10 years or brainstorm, you know, what you should
19:51do with your life or how you should unscrew up something in your life. I mean, this can
19:56be used for anything. It is amazing. But, okay, so what are we going to do here? White hat. So,
20:02the six hats, there's going to be six different colored hats. So, the first hat is the white hat.
20:07What factual elements could make your to-do app interesting from a technical data perspective? So,
20:14think about it. What kind of data could it track? What metrics, what information could it capture
20:20that most to-do apps ignore? So, I would have never asked myself that question on my own. And
20:26I'm already thinking now, wow, I was just thinking a simple to-do app from the white hat perspective.
20:30It's now getting me to think about some really interesting aspects of a to-do app. So again,
20:36there's no right or wrong answer and I don't even know if I'm going to come up with anything
20:40interesting. This is not the actual product that we're going to create. We're just throwing ideas
20:44out there. When we get the report at the end of all of this, it's going to help us distill down
20:49what is interesting to then go to the next agent with. I think it could track how much time elapses
20:56between when they enter a to-do and when they mark it done. Also, it would be interesting to
21:02correlate the length or size of to-dos and the complexity of the to-dos and how that affects
21:09how soon the to-dos get done. I love using Super Whisper, by the way, or Whisper Flow or one of the
21:15other many voice to text ones. So, I'm just going to say yes. And one nice thing about working with
21:20clawed code in the IDE is that as you work through each section of whatever whatever document you're
21:27working on, it will start creating the document for you. So you don't have to worry about being
21:33lost if you if you lose the chat, for example. I know sometimes there's anxiety when working
21:37with the web agents that you might lose it and it's hard to get the full document out, but here
21:42you're building up the document as you go. And we already have our brainstorming sessions here now.
21:47Not too worried about what's in there yet. We're just going to keep going. So now we've gone to
21:50the second hat. This is the yellow hat. Optimistic benefits. So look friends. Um this would take too
21:55long to go through all of the brainstorming. So I'm going to go through all three brainstorming
21:59techniques. All right. And look at this. We got our brainstorming session complete. We did all
22:03three exercises. Let's now look at the document that it produced. So we will go to open preview.
22:08This is actually going to give us the output and keep a record of the brainstorming that we
22:11did. So first we did six thinking hats. critical concerns, creative alternatives, and we found some
22:17fun ones in there. And then we went into process control. So, it really got me thinking about a lot
22:22of different interesting things. This led to so many different things. Then we did the five W's.
22:26The five W's was a lot of fun. The first why is why is complexity completion correlation valuable?
22:32And then I gave it an answer. And the answer was because people consistently underestimate how long
22:38complex tasks take. And the way the five W's works is then it's going to say okay well why does that
22:45or why is that the case and then I gave an answer and then it's gonna say well why is that the case
22:50why is that it's a way to really dig down in a somewhat annoying way but it's kind of fun doing
22:54it with the AI so then we get into technique three which is role playing this is one of my favorite
23:00because the AI will just invent some sort of role playing and you never know what it's going to give
23:06you so here I came up with the overwhelmed freelancer the context is You're juggling 12
23:11clients, 47 open tasks. You have a current to-do app, multiple of them, making matters worse. So,
23:19how do we triage intelligence? Show me what actually needs to happen today versus what
23:23I think needs to happen. Client workload balance, visual dashboard showing task distribution across
23:29clients to prevent overcommitment. Imagine the project that you really want to launch in the
23:34market and what this brainstorming agent can do for you. This is next effing level. Okay.
23:41And then it gave us more because I wanted to do more brainstorming. So, we didn't just start
23:45with the overland free overwhelmed freelancer. We talked about the perfectionist student and
23:51what they would get out of a to-do app, the executive assistant, the ADHD creative. So,
23:57I wouldn't have even necessarily thought of all those personas. uh the analyst Mary helped me come
24:02up with these different role playinging scenarios and we talked back and forth and then it gives us
24:06an executive summary because okay so it's great so we did all those things we're not going to use all
24:10of that right it then we kind of talked about what was important to me and I kind of like gave it
24:15some of the key insights and we came up with some of these key insights together transforming from
24:19task management to behavioral intelligence your simple to-do app becomes a personal productivity
24:24research lab that helps users understand their actual work patterns versus their assumptions
24:30Tell me that is not powerful. And that is again just from like let's brainstorm about a to-do
24:35app and this is what we came up with together. And now here's what's really cool is it breaks it down
24:40into what can we do right now? What might we want to research and do in the future and then what are
24:46like the moonshots or just the creative ambitious things we want to do later. And again this could
24:51be like any aspect of your life. So use this use this in your personal life use it for work.
24:58You can see now just how powerful the BMAD method is and you don't even need to be in the web. We're
25:03doing this all with cloud code. And as a matter of fact, this is just on the cheapest model for
25:07claude code. If I do slashmodel, you can see that I'm just using sonnet. I'm not using opus and I'm
25:14not using the default. I'm just keeping it on set. You'll learn over time which model to use.
25:19It's kind of an intuition that you'll develop. And trust me, even if you don't consider yourself a
25:24strong developer or if you're new to this, you'll pick it up. You'll understand. It's it becomes
25:30very intuitive to know which model to use. So, I would recommend though just so you can learn that.
25:35Keep it on sonnet for most things. We are done with this. We've saved everything to a document.
25:42This is the most important thing with the workflow in cloud code or really any agentic IDE that
25:47you're using. we are going to start a new task with either the same agent or a new agent. So what
25:53are we going to do? We're going to stop clear the context or if you don't want to clear the context
25:58and keep that conversation history for later, we'll just kill the window, start a new one,
26:04and restart Claude. Now, you might not know this if you haven't used Claude before. Let's say we
26:09close this and then we realize, oh, we're not done with the conversation that we were just in.
26:14If you hit slash and just hit the up arrow a few times or type slashres, you can now see the chat
26:21that we were in here before. So if you've had multiple chats, you'll see a list here.
26:25There's only one right now. If you highlight one in blue in this theme, if your theme,
26:30it might be a different color. And now we're back in the chat that we were. So that is how you get
26:35back to a previous chat. But we want to start a new chat. So again, I'm going to just close that.
26:40I could have done slashcle, but I want to retain it. So, I'm just going to do cloud again. And now
26:45we're going to start a new chat. Now, we're still going to talk to the analyst in this case. And now
26:49I want to show you how to talk to the analyst to do basically the most important thing out of the
26:55analyst that you really want is the project brief. Uh, by the way, the analyst is a totally optional
27:01thing. You don't have to do brainstorming. You don't have to do a project brief. It is something
27:05that you can do. Again, it really depends on the complexity of your project and how much you
27:09already know. If you if if we were just doing the simplest to-do app, I would go right to the PM,
27:15create a PRD for a simple to-do app. You can go right from PRD with its epics and stories directly
27:21to development. I will show all of that in the A to B full walkth through, but right now I just
27:27want to continue showing you this basic path of using cloud code. So, we're going to select number
27:31two right now. And this is now going to help us create a project brief. And you'll see right
27:35there it loaded elicitation methods. That is one of the other big awesome things that's built into
27:41the BMED method. The BMAD method is built up of YAML templates which have two things. Basically,
27:47the outline of the type of document, but more importantly, and this is what makes it, I think,
27:52more powerful than any other method I've seen so far, but other people are starting to catch
27:57on to these ideas. Embedded in the template is instructions for the LLM and how it should
28:03work with you. Because as I said a little bit ago, the beauty of the BMAD method is not that
28:09you ask it for something and it spits out a whole document, but that it really works back and forth
28:14with you coming up with the best possible thing that you guys can come up with together. So,
28:19first it's asking, do we already have any existing brainstorming results? Market research. Sometimes
28:25it's best to look out in the marketplace, see what's out there, and build your own version of
28:29it, maybe with your own twist. It's a big market out there. So don't feel like you always have to
28:34be the first to market with anything. So maybe you want to do market research and see what is out
28:39there or do a competitive analysis and see what your competitors are doing differently. But what
28:44we're going to do is we're just going to take our brainstorming. We're going to drop it here and say
28:48here's our brainstorming session results. We can use this to kickstart our work in interactive mode
28:55in creating the project brief. You don't have to use these exact words by the way. So don't don't
29:00you don't have to memorize the exact words. Talk to this as if you were talking to your business
29:04analyst. Now this is going to kickstart it. So the nice thing here too is we started a new chat,
29:08but you're seeing because we saved it to a document. I'm able to use a brand new context.
29:13It's not getting polluted. It only has the output from the brainstorming session. And so it just
29:18keeps it lean. And so now it's going to ask us some questions. It's going to pull things out
29:22of our brainstorming and probably get us through the PRD a little bit quicker than normal because
29:29you can see here, for example, it already knows what the executive summary is because we've told
29:33it when we were doing the brief. I am only going to show this first section and then I'm going to
29:38finish the document and then we'll move on to the PM. What you can see is after you complete the
29:44executive summary, this is just one part of the overall brief by the way. It already figured some
29:48stuff out from our previous brainstorming. we can correct it or we can tell it to change things. So,
29:51you'll want to read this and obviously work with it. But at the end of each section, you're going
29:56to have multiple options. These are called advanced elicitations. Whether you're working
30:02with the PM or the architect or the business analyst right here, every section of a document,
30:08most of them will ask you different advanced elicitations related to the document you're
30:12working on. Now, unlike brainstorming where it's asking you, the user questions, this is where
30:19another powerful aspect of the BMAD method kicks in, and I haven't talked a lot about this before,
30:23so this is very important, but this is where you push on the LLM to do better. When you talk
30:28to any LLM, whether you're just using the BMAD method or not, or you're just using Chat TBT,
30:33if you picture a bell curve, you know, you're kind of getting the average slice of the entire,
30:39you know, corpus of information in the LLM. It's giving you the average response. It's, you know,
30:44not necessarily great. It's not necessarily weak. The advanced solicitation is your chance
30:50to really stick the cattle prod to the agent and say, "Do better. Make sure that you have
30:55put this through the fire and you're giving me the best possible result." Challenge the scope.
30:59We can force it to brainstorm. So, just like it was having us brainstorm, we can now turn
31:03the torture back around on the LLM and have it brainstorm different ideas. There's actually
31:08many more advanced elicitations. Again, just so it doesn't overwhelm you, it's just giving some
31:13of them. Please do number four and number five. And by the way, while we're waiting for that,
31:18notice that it did already start our brief. This is our project brief. So again, the beauty of
31:23doing this in the IDE with cloud code is that we are generating the documents as we go. But first,
31:30it did critical assumption testing. So, it's basically, by the way, this is just another great
31:38thing to do with any product idea that you have. And so, I love that the LLM thought about this.
31:43Look at all of your ideas for your product and really stress test them and make sure that they
31:49make sense, especially if you want to build an MVP first to just get something to market. And then we
31:55also did explore alternative solution approaches. So at the very end of it, it kind of synthesizes
32:00everything and it gives some recommendations based on what it was doing with itself right there.
32:05Sure. This is just a demo. So we're just going to apply them all, update the section, and then we'll
32:10move on to the next section of the product brief. Let me say please uh subscribe to the channel if
32:15you're enjoying this. So much more content is coming out, my friends. You know, I know these
32:19videos don't come out as often as I would like to because I'm also constantly working on actually
32:23improving the BMED method. I'm not here to just hype up new things, new products, or, you know,
32:29even tell you I'm using this, but I do want to be able to use this forum or this platform to share
32:34the changes with you. But also, I spent a lot of time actually working on this. But subscribe to
32:39the channel, hit that notification bell so you do know when the next one comes out because it is a
32:43little sporadic sometimes. And join us in the Discord. I'm going to finish this up. We will
32:48look at the whole product brief when it's done, and then we'll move on to the product manager.
32:52All right. So, I'm going to pick up the pace here a little bit. We are done with the brief. We've
32:57gone through the whole thing. The brief is now ready for the PM handoff. So before we do that,
33:02I just want to show here's our product brief. Since we're in cloud code and not in the web, it
33:07just keeps updating the document for us. We have the executive summary, the problem statement, pain
33:13points, why we're doing this, what is the proposed solution, the target users. So this says like the
33:19demographics, who do we expect to be using this? Why would they use it? How are they each going
33:24to use it differently? And why is it valuable to them? This is so important. If you're just
33:29wanting to put any kind of product in the market and you're just thinking you're just going to vibe
33:35vibe code something, throw it out there and be an overnight millionaire. Maybe you will be, but
33:40you have to consider all these things. You have to understand the target audience that you're looking
33:45for. So that's the beauty of this product brief. It's going to really help you figure out a lot of
33:49that and make sure you're building something that makes sense or might indicate that you want to
33:54pivot on something because it's not actually going to achieve what you want to achieve. Why waste a
33:59lot of money, effort, and time building something that's not going to go anywhere when instead
34:04you can do this little bit of research up front, right? So, we're showing this this whole document.
34:08The point of the product brief is to really work back and forth with it and produce this document
34:14that helps you understand what it is you're going to build at a most fundamental level. As always,
34:20what did I say? Even in cloud code, even though we could go on and on and probably talk to three
34:24agents in a row without running out of context, we're still going to start a new one. So again,
34:30we can just kill it and start a new cloud session. I'm going to show you a different way this time
34:34and that is just slashcle and this is going to clear the conversation history and free up the
34:41context is it's it's basically the same thing. You could also do compact. I do not recommend
34:46it. And if you ever see your IDE is showing you the warning that you're almost ready to start
34:51compaction. That means that means you've been in the chat for too long and you probably want
34:56to wrap up what you're doing. There's different techniques for what you for that to get out of
35:01there, but you generally do not want to rely on compaction. It's just randomly going to forget
35:08things that might be important. So, there's other techniques, but basically, if you follow my method
35:13and you just switch and do clear, but sometimes you'll forget and all of a sudden you'll see that
35:18warning and then, you know, you have to adapt. But let's um let's talk to the next agent. So if you
35:24don't know what agent to talk to, again, you can go back to the diagram. So we've done this, right?
35:30So we we talked to the analyst. We've done our optional brainstorming. We didn't do the optional
35:35market research or competitor analysis, but you can try those on your own. It depends on what
35:38type of project, but we did create our project brief. So now we are going to talk to the product
35:46manager. And if we have a product brief, we're going to give it to the product manager. it will
35:51actually ask us. So, I mentioned you might have a similar pro or a simpler project. And if you do
35:57or you already know what you want to do, maybe you even already have a product brief, you can
36:02just talk to the PM and it will just ask you more questions if you don't provide the brief. But if
36:07you provide the if you provide the brief, it kind of gives it a kickstart and it will get the PRD
36:13done much quicker. But it just depends what you want to work. Again, if you're just doing a simple
36:17app, you don't have to go through all of this. just go right to the PM, create the PRD, even
36:22tell it you want it to be a technical PRD with architecture built into it. You'll be coding in
36:28no time. So there's multiple flows and workflows and ways you can work this system. But again,
36:33come back to this diagram if you're not sure. So let's go back to cloud code. So now again,
36:38I hit slash here. We could just type PM. That'll take us right to it. If we're not sure, we can
36:43also just kind of look through here and then arrow on it. The product manager is going to basically
36:49create a PRD which is a product requirements doc. Building a PRD is important and also I'll say the
36:54three most important things with this PRD is first of all it's going to give us all the functional
36:59and non-functional requirements and the epic that is made up to meet all of those. So right
37:06there we can see everything that is going to be encompassed by the build that we're planning here.
37:13Secondly, it's going to help work with us to figure out what is in scope for the MVP and what
37:18we can potentially pull out to have post MVP. Because if you can make your app the simplest
37:24version of the app that you can make that will meet the initial goal, that's less risk, less
37:31investment in a product. Maybe you want to get it out there with just a minimum set of features
37:35to test it in the market. That's why you want to build an MVP or minimum viable product. what
37:41is the basic thing of this core? But the PM here will also capture your other ideas and if they're
37:48not part of the MVP, we'll still hold on to those and those will be post MVP epics that you might
37:54do after the main build. So again, what is most important here? It helps us maintain MVP scope.
37:59It gives us the functional and non-functional requirements with the epics. And then the the
38:04third most important thing, but really kind of the key to the whole BMAD method is in those epics,
38:09it is going to create the user stories, but more specifically, it knows that we are creating these
38:15for very, very dumb developer agents. Now, as I've said all along with the BMED method, you cannot
38:22just check your brain out and leave it at the door and rely on it 100%. Right? So, we will use the
38:27advanced elicitation. We will read the information ourselves and logically think to yourselves,
38:33does the story sequence make sense. If you don't see a story, for example, that says scaffold and
38:41you know, set up your project or set up your accounts or whatever you need, you know there's
38:45a problem and something's been missed, right? And usually that should be the first one like project
38:48setup. You know, learn to kind of spot those things as you go. And there's also some advanced
38:52elicitations that will help us discover that. So, let's move on here. So, we are just going to do
38:56create PRD. And I want to show you before we've been doing the number, we can also just say star
39:03and we can actually just give it the command name also or you could even speak it. Uh do notice that
39:07there are two Yep. There's two different PRDs here. There's the regular PRD and then there's
39:14the brownfield PRD. Brownfield is a term that you will hear and you'll also hear green field. And I
39:21realize some people don't know what that means. So I want to explain it and it's very simple.
39:25Green field means like you're looking out on a clean green pasture. It's all open. Pretty
39:32much sky's is the limit. It's going to be easy to build your foundation. Nothing is there. Nothing
39:37in is in the way. A brownfield is where all the [ __ ] has flowed. It's a cesspool. Maybe it's an
39:45existing application that's existed for months or years or it was developed by other people. Maybe
39:52you understand it. maybe it has a lot of issues in there. If you're going to work with something like
39:57that, then you'll want to follow the path create brownfield and that will have a few more steps in
40:02it to also provided the context of the existing application that you might be working on. Number
40:08two, we're just going to do a regular PRD. Now, the PM is asking us, okay, do we have the product
40:15brief to create the PRD? So, again, this will kickstart us to the PRD. If you were starting
40:20here, you would say no and you would just start asking its question or you could go back and do
40:25it. But we're just going to drag it in because we do have the brief. And we'll say here you go,
40:31buddy. There is a lot of overlap between the product brief and some of the sections of the PRD,
40:39but it's really going to translate that from the language of the product brief into functional and
40:44non-functional requirements and a few of the other details that we need in there. Okay. So,
40:49let's see what we got here. So, it gave us some goals. So, it understands what we're trying to
40:53build, why we're building it. We got some detailed rationale here. Tradeoffs made, key assumptions,
40:59and areas needing attention. We have options to do advanced solicitation. And by the way,
41:05I love this. Just like before, as we produce our document, we are going to generate each section
41:11of the document one at a time. So, it's always going to have access to everything it's done.
41:16It's going to he keep kind of a cohesive vision of the document we're building and it's going to
41:21build it up section by section. Beautiful. I love this. Now we are doing section two. We moved on to
41:27the next section. So this is the requirements. So this is going to give us the functional and
41:30non-functional requirements. Take your time and read through this. Again, getting this right is
41:35important. And you really want to try to look, is there a functional requirement here that you do
41:40need or don't need. I want to show you my absolute favorite advanced elicitation to use here. Let's
41:46see if it suggests it. But I want to show you how you can actually find what other advanced
41:52elicitations are available for the BMED method. But what you can do is you can go to data. We're
41:58going to go to data and here are the elicitation methods. We can see them here. So there's explain
42:04reasoning, critique and refine, analyze logical flow, assess alignment with overall goals,
42:10identify potential risks, challenge from critical perspective, tree of thought, deep dives,
42:16hindsight is 2020, the if only brainstorm, agile team perspective shift. This will take
42:22your different agents and have them all look at it from a different perspective. It's a lot
42:26of fun. Uh stakeholder roundt, metaprompting analysis. So some of these are very correct.
42:32uh very creative. Um and then we get into some of the new advanced 2025 techniques that have been
42:38added in recently. Self-ont self-consistency validations, the rewoo, which is reasoning
42:45without observation. I love that. Rewoo, persona pattern hybrid, emerging collaborative discovery,
42:51red team versus blue team, innovative tournament, escape room challenge, proceed and no further
42:57action. I want to tell you, I've done a lot of research into how to really prompt engineer and
43:04get the most out of LLMs, and these are some of the best techniques that I've found from Google
43:10and Anthropic. But I've also studied multiple prompting competitions where expert prompters
43:15just compete, and I see a lot of them using some of these very cool techniques, and some of
43:21these are just crazy fun. So, let's let's pick a let's pick a wild one. I like the what if if only
43:28hindsight is 2020. So what I'm going to do is I'm going to just select this here. This is another
43:33thing you can do. Eventually these will all just be in the menu at one time. Close this now.
43:42This looks at all your functional and non-functional requirements and then it
43:46imagines that we've built the application and that we're I don't know sometimes it does six
43:50months out sometimes it does a couple years out and it's an analysis where the LLM does this
43:56conversation with itself reflecting on a board meeting or something else where people are like if
44:02only we would have done this something would have happened. So let's see what it did here. Imagine
44:08it's 6 months from now the MVP is launched. You're looking back at these requirements. What if only
44:14statement might you be making? If only we had. If only we had included basic team sharing features
44:21from the start. Non-functional requirement one might be unrealistic for bootstrap budget and
44:26single developer. What if we hadn't promised cross device sync and MVP? Functional requirement 7 adds
44:34significant technical complexity. What if we started with single device and added sync in
44:39phase 2? So what's going on here? This is actually helping you figure out how to cut scope out of
44:44your MVP. Keep it a lean lean build so you can actually get something to market. Get it tested
44:51and then layer those features in later. This is the real agile way of product development. Super
44:58powerful. So I'm going to say apply changes and update sections. Revise the requirements based on
45:03these insights. Uh before you do that, you can talk to it. You can question different things.
45:08You can tell it what you want to keep and not keep, but I'm just going to do number one for
45:12purposes of time. So, I'm going to finish this and I'm going to get this up to the epics and stories
45:17and then we'll continue together. There's only one epic right now because I told it just CRUD. So,
45:22foundation and core CRUD operation establish the project setup. As I said, always kind of look
45:27out for that, right? Uh we are going to have local database. So, it'll do the database initialization
45:32and then implement all basic to-do management commands. add, list, complete, delete, update,
45:38basic crud. So, since there's only one epic, it's going to go right into the stories here. So,
45:44epic foundation and core. You'd want to read this. And as always, read the stories and acceptance
45:49criteria and think, does it make sense? Is there something being called out that actually relies
45:56on something later? This is great. This is how agile actually works in story breakdown is each
46:01piece of functionality can kind of be its own standalone story. Can we list and search for the
46:06to-dos? Can we mark a to-do done, delete a to-do, update a to-do? You can also ask it questions if
46:14you don't see the one you want here and you don't want to look. So, I'm going to tell a deep deep
46:17think please that the order and granularity of the stories actually make sense. The sequencing must
46:24be perfect with no story dependent on a later story. It's checking to make sure that they all
46:29follow the correct progression. You're absolutely right. Let's deeply analyze the story sequencing
46:34and granularity as there are significant dependency issues in the current approach. So,
46:40do we accept it? We'll say yes. So, does this sequencing now make sense? So, let's say yes.
46:46All right. So, looks like Claude now thinks we have seven stories in perfect sequential order.
46:52Do we believe that? Well, don't check your brain at the door. Let's close that though because we're
46:56done talking to the PM for now. But you do want to, you know, go through your doc. Imagine this
47:01is a big ambitious project with multiple epics. Do you really just want to accept some slop and
47:08move on and start developing? Maybe. But I would suggest take your time and really understand and
47:14try to learn from this. Ask the LLM questions. The LLM is not going to always tell you to do this.
47:20The reason we have agents and personas is because it's to help you also have a mental model. Not
47:26not only does it help the agents perform better and keep focused on their domain expertise and
47:33you know tailoring responses to that persona that they're interacting, but it's also for
47:38you to understand that as they are serving this role, talk to them as if they were in that role.
47:46So, you're talking to a product expert or an architect or a scrum master. Ask them
47:52why they're choosing things. What do different things mean? The beauty of this too is you don't
47:58have to be embarrassed about asking questions. You can ask this what people might think is the
48:03dumbest question. Look, maybe you don't know, maybe you've heard the word database before,
48:09but you don't even know what a database is. Okay, you can Google it. It'll give you a definition,
48:15but you could also just ask why are you using a database for a to-do app. It mentioned a name of
48:21a database. You could ask it why did you select this database? Why not something else? And it'll
48:26give you an answer. So, you can learn a lot just by talking and at any time you can just
48:31ask the agent questions about this document or any of the documents we're going to do. Now,
48:37we can start a brand new chat with the architect. So, I'm going to open up Claude again. So,
48:43we're in the brand new cleared context. Nothing here. And if you're not sure, you can always clear
48:48it again just to make sure. Now, we're going to do slashchitects. Now, again, this is a very simple
48:56project. I would probably try to just go from PRD to stories. There is a workflow that supports
49:01that, but I want to show you basically how the architect works. So, same thing. It'll either show
49:07you the commands. If it doesn't, type starhelp and you can see what's available here. Now, here we
49:12have a couple different architectures. Don't be confused, but it's pretty simple. You're either
49:16going to do a full stack architecture, which is front end and backend, maybe is a monor repo,
49:24which means all the code is in one project such as Nex.js, or maybe it is two separate repos and two
49:30separate projects, React for your front end, and maybe a cloud service for your back end,
49:34and maybe in two different GitHub repos. Um, you also might be starting with a starter project such
49:41as NECJ, Nex.js with Superbase and Shad Saiyan and Tailwind. If you're doing that, you would want to
49:49generate your project, get your boiler plate and share that with the PM or the architect. So, as
49:56they're doing the architecture, they already know what you have, what your technologies are, what is
50:02in your requirements.txt if you're doing Python or your package.json JSON if you're doing TypeScript
50:08or JavaScript, Node.js, excuse me. Create full stack architectures if you're doing front-end
50:14back end. Create backend architectures if you're just building like maybe a service project or
50:19utility or anything that does not involve a web front end or a guey front end. The one exception
50:26is you might have a CLI. You will still want to select create backend architecture. And then there
50:33is a create front-end architecture. If you're only working on a web front end, maybe you're working
50:38on a website or maybe you already have a backend REST API and you just want to build the React app
50:43to call that REST API. That's where you would say create front-end architecture. And then finally,
50:49our friend the brownfield is back. So we can create a brownfield architecture. And again,
50:52that's going to involve a lot of the research that the analyst and the PM did to understand your
50:58current project. Feed that into the architecture along with some of the challenges of architecting
51:03against an existing project. Creating our backend architecture because we are doing what a to-do app
51:09only in the CLI. It's just running in the command line for demonstration purposes. So, we'll hit
51:14number two. Now, let me give you a tip on working with the architecture. Regardless of what actual
51:22IDE or system you're using, you want to use the best model you can. I'm showing you that this can
51:28be done in the $20 version of Claude Code, but I would recommend, especially on a more challenging
51:34architecture, use Opus or whatever powerful model you have because you will get potentially better
51:42results. Or do this on the web with Web Gemini. save your credits and then just get into the IDE
51:48after we have our architecture. But I actually enjoy doing the architecture here because as you
51:52can see yet again we are producing our document in real time as we go section by section. It gave
52:03us a sequence diagram. It gave us some highlevel architecture. You know what we're doing with the
52:07database, how we're going to handle error handling and global installation and uninstallation
52:12process. Do we want to focus on any of these? We can do alternative analysis, stakeholder input,
52:18risk assessment, resource impact, user experience impact, technical feasibility,
52:24market validation. You know, read what it gave you and think about if any of these make sense or if
52:29you want to do any of these. And if not, just move on to the next section. Also, question in this
52:34next section what technologies it comes up with. Make sure it knows what the latest versions are
52:40or what the best version is. You can also tell it to use the web to make sure that it's considering
52:46the latest technologies for everything that it picks is this is going to create a table of all
52:50the specific versions and technologies we're using. If you don't have this and tell it what
52:55specific technologies, what what testing framework you're using, for example, what'll happen is later
53:00on your LLM will be developing. It might be using justest just because it sees that it's already
53:05scaffolded and then it'll run into a problem and instead of trying to fix it, it might be sneaky
53:12and just try to install a whole new test framework without you noticing. It'll tell it but maybe
53:17you're not paying close enough attention or you walk away and you come back and all of a sudden
53:22there's a new test framework or there's another new package. And that's because it's not forcing
53:27it to stick to any list. When we create this table here and shard this out to a separate document,
53:32the dev agent later on will always be aware of this technology stack and will not diverge from
53:38it. It's a very powerful technique to make sure your agents are all using consistent libraries,
53:44packages, and versions. So, I'm going to go through the rest of the architecture,
53:47but just remember section by section you really want to use the advanced solicitation. So let me
53:52show you right now just one quick example example and then like I said I'm just going to kind of run
53:57through the rest and we'll pick up at the end. So let's say in here doing data models. So this is
54:01great because if it makes data models that means later on the agents they'll always conform to the
54:06same model. And much like our technology stack you don't want them producing different models on the
54:11fly. But let's say for whatever reason I don't understand why we're creating these or I don't
54:20like them. So I'm just going to ask a question. And this is one thing you can do. You don't need
54:24the prompt here always to ask just ask a question. So say explain to me the interfaces you created
54:30and I want an explanation of why each one of them exists and are we missing any? So the architect
54:36is programmed to also be very explanatory and explain in a very clear way meeting you at your
54:42level. I recommend talking to all these agents. Even if you're not in the middle of a project,
54:47you can load up these agents and just talk to them. talk at them talk to them about your
54:51ongoing project. What did he tell us? He told us a lot. So, let me explain each type script interface
54:57I created and analyze if we're missing any. So, the to-do interface that's obviously important,
55:02but why it exists? Type safety. So in other words, an interface allows us to define a object or a
55:09par or basically this would be like the blueprint for an object that we will create later and every
55:15object must have a task string completed completed at and an ID to make it unique. Um so why does
55:24it exist? Type safety. We just talked about that database mapping. So that's interesting.
55:29So what that's telling us is he's setting up an interface here that will later also map directly
55:34to the SQL light. So we always make sure we have the same properties and fields in our local
55:38database. And it's also going to help us define an API contract. The API contract when it returns a
55:44to-do is going to be guaranteed with each to-do to return these properties. So if we wanted something
55:51else in our to-dos, now would be a good time to add it, not later on when we're developing.
55:56So that's again why you want to actually pay attention and try to understand what the architect
56:00is telling you. And if you don't, just ask it. And if you read this and you still don't ask it,
56:06play five W's with it and ask it why you did this, why you did that, well then why did you do that?
56:11Well, why did you do that? And on and on and on. Again, you can learn so much just by engaging in
56:17discussions with the architect to really start to learn how software is developed in reality. It's
56:24very powerful stuff. All right. Right. And so the uh architecture looks like it is done. We finished
56:29the last section. It gave us the final summary and the document is updated. Let's just make sure that
56:36everything is in the document. It's pretty large document which is why we shard these documents,
56:40but we have our success criteria. Before we shard our documents, just like with the PRD,
56:46you really want to go through here and make sure that everything makes sense, such as
56:51coding standards. As a matter of fact, coding standards is one of one of the most important ones
56:56because this will be used by the developer agent. Directory structure. You want to make sure that
57:01this makes sense. If you want to render sequence diagrams, you can just rightclick in VS Code and
57:07you can say open preview and then you should be able to see some of these nice looking renders.
57:13If you do not have that option in VS Code or in Cursor or Windsurf or whatever you're in, you'll
57:20want to go to your extensions, which is this icon here. These are our installed extensions. And
57:26you'll want to make sure that you have something similar to Markdown all-in-one. I like this one
57:32along with Markdown preview mermaid. This will give us mermaid documents or mermaid rendering
57:39from this one. And the markdown all-in-one will also let us auto format markdown to make sure that
57:44it's valid markdown. Here's our sequence flow. Make sure that the flow makes sense. Here's our
57:52database schema. Again, if you don't understand something here, ask it. Ask it why I did this,
57:57but this makes sense to me. Here's the source tree that it's going to produce. This controls
58:03where the LLM is going to look for files, where it should put them when it creates them. And again,
58:08we'll just help it stay on the rail. Having a source tree is a very important document
58:13to keeping the LLM on track. I'm going to open up this and I'm going to show you how you can
58:18just talk to the agent. In the coding standards, I really want to make sure that the developer agent
58:23will use Gab will use good Java do style comments or JS doc style comments on all public functions,
58:32public interfaces, etc. So, please add that to the proper section in our architecture under the
58:40coding standards. Okay, that was a little bit wordy, but that's why I like using just the uh
58:48speech to text because I can kind of work it out in my mind as I'm saying it and usually the LLM
58:54can get it right and understand what I'm talking about. But this should now add some documentation
58:58standards so we will have good consistent documentation. Now, now I'm good with this. Now I
59:03think the document is great and we are done here. So now I'm going to close cloud code. I'm going to
59:09close all our documents and let's look back at our project. We now have an architecture and a
59:14PRD. The two key documents that we need. We don't really need the LLM going forward anymore. Looking
59:20at the brainstorming or the brief. Every file that you keep here is potential context that the LLM
59:26might choose to load. You can add restrictions. you can tell it not to, but it's better sometimes
59:31to add these to another folder. So, you could add these to a project references file, or you
59:36could just take them out and store them, or maybe you store them online, depending on how you're
59:41organizing your project artifacts. But whatever you do, you don't want them to get pulled in and
59:47polluting your context overall. Okay. So, we'll come back to Claude use the shard command. So,
59:54if we just type shard, we will see that there's a task shard doc. And do we want to use the MD
59:59tree? Yes, we do want to do MD tree. So, okay. So, now it's saying great, the MD tree command
1:00:03is available. If it was not available, it would suggest that you install it. So,
1:00:08now I need to know which documents you'd like to shard. Please provide the path to the document you
1:00:13want to shard. And I'll use the MD tree explode command to automatically split it into smaller
1:00:19documents based on level two sections. So, first of all, we want to do our PRD. So, I'm just going
1:00:26to drag it in there. That's it. And the default should be a folder called PRD, which it is. And
1:00:36look at that. It's already done. That's how quick it is. We now have our epic list. And here is the
1:00:42most important one. Epic 1, epic 2, epic 3 if there were multiple with all of the stories
1:00:47because this is what the scrum master is going to use to build up our stuff for our developer. So,
1:00:52that is great. So now we can say do the architecture and we're just going to drag it in.
1:01:03So as you can see using the BMAD method in cloud code is super easy. We don't have to jump back and
1:01:09forth to the web. It's super powerful and it just does a great great job at following instructions
1:01:15regardless of which model you use in cloud code. Here's our architecture again just in seconds. And
1:01:21now this is the one that is really good to shard because look we have coding standards which has
1:01:27our documentation and some of our other stuff in here that we wanted with examples source
1:01:32tree. So this shows us exactly what our project structure is going to look like. Text deck those
1:01:37three are critical and I want to show you why so you understand this regardless again of which tool
1:01:42you're using. If you go into Claude, since we're using Claude, we'll look at the Claude version.
1:01:48And let's look at what the developer agent has here. The developer agent has a command on here
1:01:54that basically tells it, look at BMAD core, look for the core config. And this is what tells the
1:02:04developer agent what files it will load every time. So the way the developer agent works,
1:02:08I've explained this before but I'll say it again. The scrum master is going to take the highle epic
1:02:14and story. It's going to read through various architecture documents. It's going to understand
1:02:19one story at a time and it's going to create a very detailed story for the developer, a lower
1:02:25level developer story. This is a self-contained file for the developer to have all of the context
1:02:34it needs to build the application. We talked a few videos ago ago about context engineering. This is
1:02:42the core of context engineering. Giving the agent exactly what it needs to build its little piece of
1:02:49the kingdom, right? But additionally, not only is all of this information the scrum master is
1:02:55going to put in there exactly what it needs from the architecture and other sources, the developer
1:03:01is always going to load these files that you have listed under dev load always files. By the way,
1:03:07I didn't have to make this myself. This file was created on the install. But if you ever wanted to
1:03:12customize it, maybe there's another file that you always want the dev to load, you can add it here.
1:03:18But again, here's our key documents that were sharded from the architecture, the coding
1:03:23standards, the tech stack, and the source tree. Take your time and really go in there and make
1:03:28updates to those if you need to and change them if any of these things change. All right, we've come
1:03:34a long way. If you're still with me, thank you for sticking with me. If you made it this far, I would
1:03:39love to know who's actually still watching. Um, because I bet this is a very long video. I hope
1:03:44people are appreciating it, though. So, if you've made it this far, how about leaving a comment
1:03:49down below and tell me what is the craziest thing that you want to brainstorm with the brainstorming
1:03:56agent. Give me some of your wildest ideas. And if you actually use the brainstorming agent to do it,
1:04:03come back and tell me one of the craziest things you came up with that you never thought you would
1:04:07come up with before. I would love to hear it. Maybe also jump into the Discord and and talk
1:04:11about your brainstorming adventures. So, as a matter of fact, some people in the Discord
1:04:15actually talked about spending hours just doing brainstorming and coming up with some amazing
1:04:19ideas. We sharted our documents. We have our architecture. We have our PRD. So, that is
1:04:26great. Now, I'm going to do something here. I'm going to create a new folder that I'm just going
1:04:32to call for now temporarily. And I'm going to say get init. put everything in there except for the
1:04:39architecture in the PRD sharded documents. And now I'm going to add agit ignore. So I'm going to say
1:04:46new file.getit ignore. I've added agit ignore file. Please populate it with the common things
1:04:54that should be in a git ignore and also include the ignore folder in the git ignore and the.getit
1:05:01get ignore those files should also generally be ignored by clouds so it won't clutter the context.
1:05:08So there we go. So we just populated a get ignore with a lot of common properties. Now if you're
1:05:17using a starter template or a starter project, it's probably going to set up your git ignore
1:05:21for you. Or if you're using like Nex.js, it'll it'll set that up for you. In a complex project,
1:05:26you would want to use the PO at this point. And the PO has a command called run checklist. And
1:05:33it's going to run the PO checklist. What that does is it's going to look at all of your user stories
1:05:39fresh and it's going to look at the architecture, make sure that they're aligned, make sure that
1:05:46nothing came up in the architecture that needs to feed back into the stories because maybe something
1:05:51changed or if there's any big gaps. It's a really good idea to do that. Sometimes it'll find things,
1:05:57but this is such a simple app and like all of the BMED method, it's adaptable. We do not have to do
1:06:03that step. So, we're going to skip that and we're going to jump right to the scrum master. Now,
1:06:10if you've never used the scrum master, you might not know what it can do. So, as always,
1:06:13I recommend doing /help if it doesn't give you the help commands automatically. Sometimes it
1:06:19will. That's just the nature of LLM sometimes. Um, so we have a few things here. We can draft,
1:06:26which means create the next story. We can do correct course. This is a feature that
1:06:31a lot of people have asked for and don't know exists. Why does my chair keep sinking? Um,
1:06:36correct course is something let's say you've developed, you know, halfway through some of
1:06:40your epics. You're mid project in the middle of a story and you realize you forgot something or
1:06:46you want to make a big change. Wrap up what you're doing and then talk to the scrum master or the PM.
1:06:52You can choose either one and run this correct course command. It's going to ask you a bunch
1:06:57of questions and it's going to analyze how far you've gone and whether it's better to, you know,
1:07:04revert back to a certain earlier stage and produce new stories and epics or maybe generate new future
1:07:11stories and epics from where you're at right now. Or your change might be so drastic that it just
1:07:16recommends, although it generally shouldn't, that you start over. But not only is it going to give
1:07:22you these suggestions, it is going to help you. It's going to figure out what epics and stories
1:07:27need to be modified or changed or added or removed so you can do this pivot, account for it, and
1:07:33still get to your end goal. Maybe it means making updates to the architecture or as I said the PRD
1:07:40or maybe you found a new API or a new library that you want to do use, so it might put it in the
1:07:45tech stack. All of that's a possibility with the correct course command. That could be a lifesaver
1:07:50depending on the situation. Of course, it's best to avoid that which is why you go through
1:07:54a lot of planning. What we're looking for here is number one, we want to draft the first story. Now,
1:08:01if you know what story it's time to draft, just tell it. We know we want to do epic one,
1:08:08story one. So, what we can do is we can just say we can say we can say star draft 1.1.
1:08:16So, let's say we're halfway through and it's time to do story 2.3 because story 2.2 finished. You
1:08:23can just tell it draft the next story and it will figure out what the next story is. Or you can tell
1:08:28it which one to do. That'll just make it a little bit more streamlined. It'll go faster and it'll
1:08:34start drafting drafting the story that it needs to. When it drafts a story, it's always going
1:08:39to check if there was a previous story. if there was, it will also check if there's any notes in
1:08:45that story that it needs to carry forward into the new one. A lot of times also these project
1:08:50foundations will have steps in them. So, you really want to read these stories especially and
1:08:54see if there's any human basically things that you need to do. Maybe you need to set up an account or
1:09:01maybe you need to, you know, visit a website and provision some infrastructure that for whatever
1:09:07reason cannot be done by the LLM through a command to the remote service. Maybe you have to get out a
1:09:14credit card and pay for something. Okay, so it's done. Let's look at our story and let's see if
1:09:18it followed the proper template. And it's always going to put it in draft mode. This is standard
1:09:23agile story practice. So as a developer, I want to establish the basic noode.js JS TypeScript project
1:09:29structure so that I have a solid foundation for CLI development. Acceptance criteria means what
1:09:35does it actually take to say yes this story is done and complete. The tasks and the subtasks
1:09:42are going to sound a little bit similar to this a lot of times and it should because these are the
1:09:48actual steps that the developer agent is going to follow step by step to implement our story.
1:09:55This is where the scrum master agent goes through all of the architecture documents,
1:10:01finds information that is relevant to the story and gives it to it here. Since we're
1:10:06doing project setup, the agent needs to know the structure of our project that the architect came
1:10:11up with. So that's why it pasted it here. So the developer agent has this contextual information
1:10:16and doesn't have to search for itself bloating its own context. So there are placeholders where
1:10:22the dev agent can put notes and also where the Q agent QA agent can put their notes. Change this to
1:10:30approved. We are done with the scrum master. We're just going to close our context and
1:10:35we're going to start a new one because as always between every step even though we don't have to,
1:10:41it's better practice and it minimizes the overall context. All right. And now we're here in the home
1:10:46stretch. We got James, our full stack developer is locked and loaded. Let's do star help. Now you can
1:10:56just tell James develop a story and it will find the highest numbered story that is set to approved
1:11:02or already in progress that maybe wasn't finished by an instance of himself from before. Or you can
1:11:08just tell him. So why don't we just tell him then he's not having to do the searching. It will just
1:11:12be more uh more streamlined. So we'll say star develop story. We could have also just told it
1:11:23develop story. We could also just say number five. Regardless, I'm just going to hit space. And then
1:11:29I'm going to drag in the file. And now we should finally see our developer implementing story one.
1:11:36I figure let's just do the first story and then we'll commit. But we're just going to say yes,
1:11:41let it do whatever it wants to do. It's setting up our package.json. JSON. And one thing we can
1:11:46do is as it adds the different packages that we need as it's going through the different subtasks.
1:11:51We'll make sure that this is aligned. By the way, we could turn on the unsafe mode for cloud and it
1:11:57would just churn through. But for right now, I just have it. I'm sitting here. We're doing
1:12:02this together. I don't mind just telling it yes, continue. If you want to roll the dice,
1:12:06you can turn off all the safety features and let Claude go wild. And look, we're
1:12:11uh we're almost done. It it created a bunch of its own internal to-dos mapped to the actual
1:12:17to-dos in the story. And then we should see that it will actually go through. Now it's starting to
1:12:23check them off. We'll set the status to ready for review. So when it does that, you have two
1:12:28options. I actually recommend doing both, but it depends on the story. If it's a simple story,
1:12:32probably not necessary, but review it yourself. Maybe test out the functionality manually. Check
1:12:38it out. I would start a new context yet again. So, we're just going to exit. We're going to say
1:12:42Claude. And now we're going to use our QA agent, which is quality assurance. So, let's do star
1:12:49help on Quinn. Quinn the QA agent. So, let's do number one. And let's pull in story number one.
1:13:00Now, again, this is a fresh, brand new context. So, it's going to be looking at the story and all
1:13:05of the project source code and other updates that it made to really make sure that it did
1:13:12a good job. So, this is going to this is going to do a pretty good deep examination.
1:13:16This is another place where you might want to use Opus. Maybe even more importantly
1:13:23than when we were doing the development because this is the critical kind of piece that makes
1:13:28sure that the agent didn't go off the rails and stick files in a stupid place. Um, and it
1:13:34looks like it's already uh finding some things. This was obviously a very small setup story. So
1:13:39there's not a lot for the QA to do for the QA to do, but it did find a few things and it gave
1:13:45us uh these compliance checks that it did here or improvements that it made. So this is great,
1:13:49right? I just want to say thank you again. If you stuck all the way through this video and
1:13:53made it through here, you're going to be so far ahead of the game. You're going to be a
1:13:58power user using the BMED and understanding the BMAD method, especially in claude code.
1:14:03I welcome you to come into the Discord forums and give me your shared experience of how it's
1:14:10working for you. Share with the community or if you have questions still ask the community
1:14:15and they'll be happy to help you. I'm sure also it's a great it's a great place. We'll see you
1:14:19next time. My name is Brian. This has been a very long episode on the BMAD code, but I hope
1:14:25it was helpful and also worth waiting for. So, thank you everybody and we'll see you next time.

About this transcript

This page contains the full transcript of The Official BMad-Method Masterclass (The Complete IDE Workflow) by BMad Code, generated from the public captions YouTube serves with the video. The transcript has 14,441 words across 796 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to translate the transcript into 100+ languages.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, translation to 100+ languages, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.