Transcription
In this video, I'm going to show you my personal development workflow on how I be able to build features on an existing projects using five to 10 sub agents inside of claw code. So, it doesn't matter if you are in a big tech company or working at a startups, chances are you're probably going to touch with existing codebase.
So, that's why in this video I'm going to take you with me to explore how I be able to use different sub agents like five to 10 different sub agents inside my current development workflows and how I be able to use different sub agents for different types of task. So, I do know that this is not the first video that I made on sub agents. So, if you're interested to learn more about how to use sub agents, you can check out this video right here.
But for this video specifically, I'm going to show you my curated sub agent list on how I use it inside of my day-to-day workflow. And also if you're curious on how I be able to use different type of specialist agents that I made video on for example like supercloud or bat method even the spec driven development from the spec kit from GitHub you can also comment down below and I'll try to use or do a demo video on how I be able to use those type of frameworks inside of my existing projects or existing codebase.
So with that being said if you're interested let's get into it. So basically the workflow usually goes like this where I basically have meetings taking place with different people like the product managers or stakeholders or maybe cross teams try to understand what we're trying to fix. We're trying to understand what we're trying to build for this new implementation. Right?
So basically I have this transcript feature turned on for my meeting. So if you're using like third party applications, you basically have the options to basically enable the meeting notes or a notetaker to basically transcribe the meeting into plain text and I can be able to take that documentation for meeting notes and be able to pass it to claw code and summarize that meeting notes and be able to gather the requirements for the projects based on that. Right.
After I gather the project requirements, I'll basically have a sub agent here called the product strategy advisor, which I'll about to show you in a second or you can also found in the link in the description for the full sub agents. And basically, I have it to summarize the project we're trying to do. Basically, I first provide the context on the existing project that I have implemented and also the next project we're trying to work on as well as what are the things that I want the sub agent here to do to basically rename the files to basically review the documentations and make some improvements and also make sure that it's developer friendly so that it's much easier for other people to understand.
So once we have our plan agents then we also have our system architect agents and that's responsible for creating a plan or technical plan on how the developers is going to execute the plan to complete the project.
Now, a quick break to tell us about the sponsor of this video, Alskill. Remember when people were worried about AI will replace their jobs? Well, the reality is it's more about who can use AI more effectively. But here's the good news. You can use Alskill here to improve your career to get ahead of the AI curve. And LS skill here is offering free access to their 2-day AI mastermind from Saturday and Sunday 10:00 a.m. to 7:00 p.m. But usually, a program like this will cost anywhere around $895. But for the first a thousand people who sign up through my channel, it's completely free. With 16 hours of live training, you get hands-on training with 10 different AI tools. From building your first AI agents to automating your work with AI automations, as well as building apps and websites with AI, with over 4 million people have gone through this training, you will get access to bonus that's worth over $5,000 if you join both days. Grab the free seats now using the link in the description below. Or you can also scan this QR code to join.
So here you can see I list out the projects or the past projects that we have completed. And here you can see that these are the requirements that we're going to do for the technical doc right. So we're going to create a implementation task list. So what are the things we're going to break down. So for example like specific task the measurable task and also things like the check upon the completion. So what so we have a checkpoint for each uh task and also the size of the task as well. And here you can see these are the task list formats. So basically you can see that after I have submitted this prompt here you can see that the system architect has executed and created a entire task list for different phases of this projects. So all the way to five or six phases and here you can see there are tasks that can be run in parallel which can be run you know at the same time right and then here you can see after I have the uh technical doc is created then I have another sub agent called a staff engineer here try to do the prioritization on the task list. So what are the tasks that we need to implement or that need to start right away right? will tell me what task we should start and here you can see it gives you a response uh here you know here are the task that we need to prioritize to start first so you can see a list out you know what we need to complete for day one and here you can see 1.1 to 1.3 and then here you can also mention that hey I wanted you to start with 1.1 for staff engineer sub agents and here you can see they start to do that and here you can see the staff engineer execute 1.1 task which is to check the Azure open AAI API validation which is the task one to making sure that the the API response successfully and the performance must exceed the target right by simply making a curl call to making sure that he's able to respond in the right time and also making sure that we uh set the budget limits as well as the security air handling and monitoring.
All right, so after that I also have the staff agent here or staff engineer start move on to the task 1.2 and here you can see that it has complete the task for 1.2. Then after usually one task is completed I usually have another sub agent here called the senior co-reviewer agent. Basically the goal is to review the documentations as well as the code changes that is being made and be able to provide feedbacks and that are actionable that the other AI agent here can be actually follow to execute the feedbacks. So here you can see that I'm just going to have this senior code reviewer here to review the task 1.1 and two and be able to review them update status and provide the feedback after reviewing it.
Awesome. So now you can see that after reviewing everything you can see that here it gives you a score as well as is approved for the phase three continuation and here you can see that these are the key finding for what are the things that are good and here are the minor improvements for the recommendations for example integrated for the production developments uh deployments as well as the enhance the security with the API rotations unit test coverage and such. So after we looked at the recommendations if we're okay with the current status then we can just proceed with it and come back with the minor improvements if we have any. So here I'm just going to proceed with the step three. So proceed with 1.3 for the staff engineer here. So I'm just going to enter this and proceed. And then you can see that the 1.3 task are fully complete. Here you can see it gives you the approved rating from the senior reviewer. So it's probably time to review the task 1.3 as well. So I'm just going to have the same prompt here to basically do the same for task 1.3. And please review it and update it status as well. So I'm just going to have the senior engineer here to review it. So pretty much you can see that from now on it's just going to be repeating this process to have the staff engineer here to implement the code and also have a senior code reviewer here to basically review the code that the staff engineer wrote. Right? So basically you can be able to reuse the same prompts over and over again by changing the parameter for task 1.3 to the next task and be able to have this complete 1.1. Right?
All right. So now in terms of the sub agent that I just demonstrate, I basically use the sub agents that I create personally. And here you can see that these are our sub agents that are refined. First sub agent that we use is the product strategy advisor. Basically the goal here is to specialize and analyze the codebase try to understand what we're trying to build for this product. So I basically take the plan the markdown format that I just show you and basically taking the sub agent that I have and try to understand and try to have it to understand what we're trying to build first. Right? So it's going to create a plan for that.
And then if we were to go to technical which is step two to use a system architect which is another sub agent that I built. And really the goal for the sub agent here is to create a scalable and maintainable system plan that the developer can follow to implement the project. So it's basically going to create that plan into an actionable plan or actionable task list that developer can follow to basically proceed with that plan and complete it one by one.
And then later on if we were to execute this plan we also have a staff engineer which is another sub agent that it creates. So which is this one. And then basically what it does here is that it's going to focus on the full stack development. It has basically the core knowledge on different languages on different database right different architecture and most importantly it start with knowing or understand the task first before it start to do the implementation and then after it start to do the implementation right we will also have another sub agent here called the senior code reviewer which the goal here is to review the code that the previous agent wrote right focus on doing the code review making sure that the future that we have built is fully is fully complete right and basically You can see that for the remaining of the demo, I basically try to continue to use the staff engineer and then also the senior code reviewer here to basically cycle through and be able to execute the task one by one.
And just like I mentioned in a couple videos ago, using a sub agents, you can be able to improve your context window for your claw sessions. But obviously, nothing is perfect, right? So these are sub agents that I use personally and I continuously try to refine those prompts that we have here based on the project we're trying to build. But these are just basically some example prompts or some example sub agents. So I would say that if you have like better prompts so basically I have this project or this repository committed and you can check out the link in the description but if you have a better idea on how to can refine this feel free to commit or make a pull request uh be able to build on top of this so that other people can use it.
So typically that's basically the workflow that I follow for using different sub agents on an existing project that we're trying to build.