Bob Martin Releases Tool for Controlling Code Generated by AI Agents

AI agents still require high-level control, but manual code review is becoming a bottleneck, and the capabilities of modern models mean that rigid deterministic verification systems are starting to limit their potential.

This raises the question: how do we control the results of their work?

To address this, Bob Martin, known as the author of "Clean Code", created a special system analysis tool integrated with an agent. It displays the project structure as a UML-like diagram and highlights it with CRAP and mutation testing metrics.

The diagram is interactive. You can select any element and gradually drill down from the overall architecture to a specific code snippet.

If a user sees a problematic part of the system, they can immediately instruct the connected agent to make changes. This could range from architectural restructuring to local refactoring. The agent applies the edits, and the updated structure is then reflected on the diagram.

He built the tool for his own needs, but it can be used as a template, and you can instruct an agent to build a custom version for your specific project.

It is currently a working prototype, so it will likely change significantly in the coming weeks. However, by then, you will probably already be adapting it to your own workflow.

Improvements can always be pulled from the source or from other developers' forks and integrated into your version.

github.com/unclebob/uml-viewer