Blog

Posts

Doesn't a Fist Have Five Fingers Already?

Testing the app structure that Claude Code created.

Tags: Agentic AI, Claude Sonnet 4, Claude Code

Sunday, August 3, 2025 9:29 PM

Last Updated: Sunday, August 3, 2025 9:54 PM

This post is part of a series. Follow along with the development progress here: moodring-vibe repository. More to come!

Please note that this is not a tutorial. This is a documented first-time experience with a fully-enabled AI agent. Do not take this as professional advice. Always protect your data, use version control, and refer to documentation if you decide to follow any of these steps yourself.

We've already set up our database, back end, and front end by letting Claude Code follow our app spec. Now, we can start tying things together and making sure things are working the way we intended.

Self-documenting in a different way

Before we begin, I noticed that there was a warning in my editor. Let's have Claude fix that. We'll set up some rules shortly that ensure these things are checked more thoroughly before any commit.

I've already mentioned that agentic coding gives you much more opportunity for test coverage. This is also true for documentation. If Claude can tell us how to start our app, it can also populate the README for the repository.

Accountability

We have already specified that Claude should commit and push any changes after each prompt is completed, but it seems that wasn't enough.

And we want to make sure our agent checks for lint errors and warnings, such as the unused import warning we just encountered. These things should be fixed before each commit.

We have changes in our repository! We can start working.

The back end starts fine using Claude's instructions. But we missed an important step in front-end setup.

We're up! Let's make sure our simulator is working.

"Mr. Watson, come here"

Before we start building out our actual app, we want to make sure the back end is serving data.

And since we're just testing things, we don't want to keep this code in the codebase when we're done.

Let's make sure the connection between the database and back end is working properly.

We also want to make sure we can delete rows from the database.

We can verify that Claude isn't tricking us and only mocking the data in the front end...

...and we can see that the data is being udpated in the database.

Now we can start getting into the meat of our actual app.

Writing an employee handbook

At the end of our previous post, we asked Claude to be more critical of things we ask it to do. Since that change occurred while writing the blog itself, it happened after all of the above activity. Here's the prompt I used to to update README.md.

And finally, the blog is caught up with what I had accomplished with Claude Code in the span of an hour or two. Going forward, we'll talk about what I've learned since then—how to better use the Claude Code CLI and the different use cases for subagents and MCP servers.