Transcript
[00:00] I have done a handful of videos on GitHub SpecKit in the past, and somebody had a really good question the other day, and that was, do you have a preferred tool for deriving specifications? Now, if you’re not familiar with SpecKit at this point, it is a toolkit by which you can start with a loosely defined specification for a feature or an application, and then run that through a series of commands that begin with a proper specification through task identification, and then implementation, and then finally validation. And the short answer to the question is no, I don’t have a preferred tool for deriving specifications.
[00:42] However, I do have a tool I can highly recommend, and that is the SpecKit Assess Extension. So with the Assess Extension, we get five new commands, and I’m going to run through them really quickly here. The first one is Intake. You give it a very basic statement about the application or feature you’re trying to develop, and it kicks off this whole process. The next one is Research, and this is going to do a lot of research, just like it sounds. It’s going to look at other applications or features within other applications that are similar to the one you have. It’s going to gather supporting and opposing information as they relate to the feature or application you’re developing.
[01:22] Then we move on to the Define command, which really should be called the Problem Statement command, because that’s exactly what it does. It looks at the users of your application and tries to figure out what problem it might be solving for those users, what gap it might be filling, and what the success criteria would look like in order to fill that gap for those users or solve that problem. Then we get to Shape, which is going to look at your existing code base and start evaluating the tasks that are going to need to be done, the architecture and how it’s going to be impacted, and the level of effort in order to produce this feature or application. And lastly, we get to Decide, which is the final command in this whole flow, and you’re going to get one of three results from that, and that’s going to be Go, which means everything’s great. Let’s just pass this off to the remainder of the spec kit processes. It could be Needs Clarification, which is going to ask you a series of questions and then jump back to a previous step and move forward through the process again. Or it can come back with Kill, which basically means this is a bad idea. The level of effort isn’t worth it, or it may just go against our spec kit constitution, which we established when we first started the application.
[02:41] So if you don’t already have spec kit installed, you can do that with UV tool install specify CLI. Now I’ve already got that installed, so I’m not going to worry about that. And if you are beginning a new project, you simply say specify init, and then the application where your project is going to live, you choose your agent, and you’re all good. Now I’ve already got a project up and running, so I’m going to switch to that. So to install the spec kit assess extension, we’re going to run specify extension add assess. And cool, that has been installed again in our application where we’re already using spec kit.
[03:40] I’m going to go in and open up open code. I’m going to forward slash spec kit dot assess. And we can see we have all five of our commands here, intake, research, define shape, and decide. So let’s take a quick look at our code base. You can see in this application, I’ve been using spec kit the entire time. So I’ve got all these different specifications and the documents that relate to those various specifications. Now in our spot specify directory, I now have this new directory called extensions. And in the commands of assess there, I have the five commands or skills related to the assess extension.
[04:07] Now we’ll take a quick look at the application itself. It’s very, very simple. It’s basically a task manager. I can come in here and add a new task with a title, a description, a priority, a status, and a due date. I can save that guy and now it’s in the list. And let’s see, we’ve got a edit. So we can come in and change some things. We’ve got a calendar view and I set that for October 21st. So there it is. And I can open it up and edit it from here as well. Super simple.
[04:42] So let’s kick off our assess process. I’m going to jump in here to open code and we’re going to run spec kit dot assess dot intake. Now, slightly against the norm, I’m just going to run this command all by itself. In most agents, you can pass in the statement and the slug right after that. But open code is a little weird about it. So I run it naked and then it asks me for those inputs. Cool. So here we go. I am going to drop this in and I’m basically saying add a sort in addition to the filters that we already have in the application. And now I’m going to find a slug and that slug is going to be list order. And the slug is just something we can reference later. Now looking at the application, yeah, up here I’ve got filters, but I don’t have any way to sort these guys. So that’s what we’re asking for. So we’re going to go ahead and run that. And again, this is the phase at which the process is going to capture our initial idea and set up our initial documents.
[05:40] Once that wraps up, we are going to run our research tasks. So spec kit dot assess research. And then we’re going to pass in our slug. So slug equals list order. Cool. So we’ll let that process run. And now that that’s done, we’re going to grab this define task. So spec kit, assign, define, pass in our slug. And we’re going to run that guy as well.
[06:13] And while that’s running, we’ll take a look in this new assessment directory where we have our intake document. It lists a bunch of clarifications and the overall kind of idea behind our feature. And then we have a research document. And if we look through this guy, we can see prior art, marketing and context, data and constraints, all sorts of research that it’s gone and done on our behalf in order to facilitate this task. Cool. Now that our define process is completed, we should have a problem doc. We’ll go ahead and jump at the application. And there it is. There’s our problem doc. Uh, it looks at affected users and stakeholders, what our goals are, what our non-goals are, and again, our success criteria or, um, metrics.
[07:04] So we are going to move on to the shape process. Again, this is going to create the general concept as it relates to our architecture, our code base, uh, and what the level of effort is going to be behind this feature. And now that that guy’s wrapped up, we’ll take another quick look at our documents. And sure enough, we have our concept MD, we can scrub through this. It is looking at, uh, all the various aspects of our code base that are going to be required, uh, to be touched. And it’s also come up with a handful of options. So we’ve got an option A, minimal sort, uh, option B sort with persisted preference and option C of do nothing and, uh, and then recommendations. And in this case, it is recommending option A, which is a minimal sort. Um, it also lists some elements that are out of scope and some assumptions that it still needs to validate.
[08:02] So with that, we are going to jump into the decide flow. Is this a go or a kill? So we’ll say spec kit dot assess dot decide. We’ll pass in our slug once again. And we’ll see what, uh, what the agent says. And we can see here, it came back with a decision of go. And we can see here, it came back with a decision of go. The idea is worth specifying. It’s recommending the option A that we looked at earlier. And it’s referencing the fact that it did in fact, create a decision document. We’ll take a quick look at that. There we go. And it’s basically everything that it said here with all of the, uh, backing or supporting information available to us in this Markdown document.
[08:56] Now, as far as the specification, everything after this safe go line is our specification. Now we can just copy that. And if we want, we could simply say spec kit dot specify. And paste that in as our specification. However, rather than doing that, I am going to clear this out and I’m going to jump into a new. Open code session. And from here, I’m going to say spec kit dot specify. I’m going to say, use the handoff summary in. I’m going to grab that document. Then I’m going to say to specify the approved concept. And this is going to kick off our entire specified process without me having to copy that or paste it or even touch it. Uh, and once that gets wrapped up, we’re going to do our spec kit plan. We’ll run that guy. And now that that’s wrapped up, we’re going to run our spec kit task to implement or create the list of tasks that we need to complete. And with that done, we’re going to run our spec kit dot implement. And when that guy finishes, we are for good measure going to run spec kit converge in order to challenge that implementation. And once that guy’s wrapped up and there are no challenges, I can jump into our application. And sure enough, I can see with a refresh, we now have our sort by priority, by title, by due date. Exactly as we defined it.