Transcription
Did you know that GitHub just released an impressive new open-source toolkit that empowers developers to use your AI tool of choice? But that's GitHub Copilot, Cloud Code, Gemini CLI, and many others for spec-driven development. It's called SpecKit.
Instead of relying on Vive Coding, where you can just throw a prompt to an AI agent and hope the code works, SpecKit makes the specification the central artifact of your workflow. Specs become living executable documents that guide your coding agent, evolve with your project, and keep everything aligned.
But why would you even use SpecKit? Well, this is an open-source tool that is under the MIT license and it is fully hosted on GitHub by centering your development process around specifications. It essentially works in a four-phase process where you have different checkpoints. You first start off with the specification where you define the goals and the user outcomes. The planning phase where it maps stacks, architecture, and constraints. You have the task phase which is where it breaks it down into small, testable units. And lastly, is the implementation where the AI codes step by step for review. And overall, with all of this being taken into account, specs become living artifacts, which is ensuring better alignment, reliability, and higher quality AI-powered development.
Also, guys, I've been posting a lot more on my second channel, and this is where my team and I have been trying to focus on a lot of news topics. So, if you're interested, go ahead and take a look at this with the link in the description below, 'cause there is a lot of information that you can gain from these different videos.
To get started, it is really simple. Head over to the SpecKit toolkit on GitHub, which I'll leave a link to in the description below. And essentially, it's a CLI tool that you can work with called `specify`. You can initialize it to your project depending on your coding agent you're using. So if you're using something like Cloud Code, you can initialize it to that with this command over here. You can simply go ahead and copy this portion and then paste in your project name. So, for example, I can paste in this command that was given from GitHub and then I can specify my project name and click enter.
So you can see right here that it has created the new project based off of the name that I've given it and it is going to request me to choose my AI assistant. So I have the option to use GitHub Copilot, Cloud Code, as well as Gemini CLI, and I can even use another coding agent which I would have to configure in the back end. But what I recommend you use this with is obviously GitHub Copilot. But Cloud Code works really well with `specify`. So I can go ahead and click enter and it's going to initialize the `specify` project. This is the project name that I have. It's going to have it so that it checks the required tools, selects the AI assistant, which is Claude, fetches the latest release, downloads the template, and essentially just initializes that overall project.
Now, what it recommends that I do is head over to the `specify` project, but it also recommends that I open it up within my IDE, and I can use the slash commands with Cloud Code. This is where you can type in the slash command in any file to see available commands specified to create specifications. The `/plan` command to create the implementation plan, and the `/task` command to generate the task. And then it is going to update the `constitution.mmd` file within your project's non-negotiable principles. And essentially, that is how you can get started with spec-driven development. Like we had mentioned before, it works in a four-step process where you specify, plan, task, and implement.
So I've opened this up within VS Code. I can open up a new terminal and we can follow the steps that it has told us. So what I can do now is within my `specify` project folder with Cloud Code initialized, I can now use any of the slash commands with SpecKit. So what I can do now is use these different slash commands like `/plan`, I can specify, use the `/task` command, as well as implementation. Start off, whatever you're trying to do, you want to first create the spec. You want to specify the command to describe what you want to build and focus on what and why, not just the tech stack. So in this case, they recommend that you specify, use the `/sp specify` command. And you can paste in whatever prompt you want. So I'm just going to use the example that they have. And I'm specifying that I want to build an application that can build and organize my photos in a separate photo album. And you can see that it is going to create the specifications for the AI to follow.
Next is where we're going to then drop it off to the planning phase. So in this case, it will use minimal number of libraries, use vanilla HTML, CSS, etc. So the second phase is the technical implementation phase with the planning. So after it finishes creating the spec file, we can then use the `/plan` command to provide the tech stack and the architecture. So it looks like it has finished generating the specifications where it gives a rough guideline on what we need to do to create this. So now we can actually take a look at this. So if you click on specifications, if you open the preview up, it'll give you a rundown on the specifications that the AI agent will follow.
Now what we can do is go on to the next phase. This is the planning phase. So if you scroll all the way to the bottom, we can paste in the next command, which is `/pl`, which is where we're going to have it now generate the implementation roadmap for all the features it has specified within the specification MD file.
Before we get started, I just want to mention that you should definitely go ahead and subscribe to the World of AI newsletter. I'm constantly posting different newsletters on a weekly basis. So, this is where you can easily get up-to-date knowledge about what is happening in the AI space. So, definitely go ahead and subscribe as this is completely for free.
So you can see right now that it will also go ahead and configure all the different templates that will be necessary for the AI agent to follow. It'll also focus on creating all the different scripts to use, as well as keeping a memory of everything that was generated.
Now what I can do is have it ready for implementation where I can run the `/task` command to command the generated order implementation task following the different principles that were basically outlined with the different lining phases. So essentially now it's going to create the actionable list and it's going to have the agent break it down and implement it. So it will then work on creating it and now it will thoroughly focus on following the each different specifications that were listed out from the planning and specification phase and it's going to implement all of it.
So rather than going ahead and having an AI coding agent take your prompt and developing a full-on application which probably has a lot of errors, it hallucinates and doesn't generate almost perfectly. In this case, with the spec-driven workflow, it is going to be able to thoroughly work on iterating on your project idea. With the planning phase, which you can iterate on further and add more specifications, it has less chances to hallucinate, uses less tokens in my opinion. You can see quite quickly it was able to focus on developing all the different features that we wanted a part of our photo album organizer. You can upload different thumbnails like the SpecKit thumbnail over here. You can upload the photo to the different uh the albums that you have uploaded or created. And you also have the ability to manage all these different albums within the main tab. This is a really simple app that lets you organize your different photo albums. But more importantly, it was able to follow the structure as well as focus on the back end as well as the front-end system. So, it is going to be able to thoroughly develop your app with all the features that are actually functional and gets the job done with higher quality.
If you like this video and would love to support the channel, you can consider donating to my channel through the Super Thanks option below. Or you can consider joining our private Discord where you can access multiple subscriptions to different AI tools for free on a monthly basis, plus daily AI news and exclusive content, plus a lot more.
Now, there's a lot more to this and it's recommended that you actually take a look at the documentation to understand the best sort of output you can get with spec-driven development because there's a lot more to it when it comes to actually prompting with these different phases. So, I definitely recommend that you read through this so that you get a better understanding of how you can approach these different phases to get the best results. I'll leave a link to this GitHub repository in the description below.
But with that thought, guys, I hope you enjoyed today's video and got some sort of value. This is something that you can easily get started with today and easily build high-quality apps or just generate high-quality code. But with that thought, guys, thank you guys so much for watching. Make sure you go ahead and subscribe to the second channel if you haven't already. Make sure you join the newsletter, join the private Discord, follow me on Twitter, and lastly, make sure you guys subscribe, turn on the notification bell, like this video, and please take a look at our previous videos because there is a lot of content that you will definitely benefit from. But with that thought, guys, have an amazing day, spread positivity, and I'll see you guys fairly shortly. He suffers.