Automated AI Feature Dcoumentation

Tracking development history

The Problem

Thanks to AI coding agents, features can be built quickly, code can be refactored easily, and it lets us have more time to spend thinking.

The problem is that, even if you take the time to review the code, you do not recall well what was implemented and how.

I found two reasons for that:

• You ship many many more features than before; your brain is not yet trained to recall them.
• You did not spend time writing the code, so your brain has seen only a very short snapshot of it.

One way to keep track of what has been done is to look at commits or PRs.
However, I work alone on my projects; I make a commit whenever I want, sometimes a dummy commit like “fixed this typo”.
I commit at the end of the day—sometimes I did a lot, sometimes not—so that is not reliable.

Also, I noticed with GitHub Copilot Agent a “default behavior” which is: add documentation anywhere each time you do something.
This led to a mess in my repo, with markdown pages everywhere.

Solution

My solution is to ask the agent for incremental feature notes.

In your AGENTS.md or CLAUDE.md, add the following:

# Documentation

Whenever you finish **developing** a new feature, create a documentation note in the `docs/` folder.
Notes are indexed as `000-my-feature-name.md`, and the number is incremented each time you add a new feature.
In this note, explain the goal of the feature, what **value** it brings.
List which part of the code has been modified, and what the main changes are.
Add any relevant information.

Customize this as you wish based on your needs.

The idea is to have the agent writing a debrief note 123-meaningful-name.md so you can see:

• the history of what has been done
• get a detailed view of a feature.

>> You can subscribe to my mailing list here for a monthly update. <<