Cleaning Up a Repo Isn't the Same as Preventing the Next Mess

I spent part of today cleaning up my article-pipeline repo.

One decision record contained a claim I couldn't verify from anything in the repo, so I removed it. Three small driver scripts had no tests, imports, workflow references or documentation pointing to them. One was already marked stale. I deleted all three; git history still has them if they're ever needed.

I also rewrote the README. It described some documentation as living outside the repo and listed pipeline stages that hadn't actually been built.

While doing that, I added a rule for future changes: every new script, directory or integration must be identified as either reusable product code or installation-specific, with a reason for the choice.

I'm not sure yet whether this will prevent much future cleanup. It might just add another rule to maintain.

But fixing what's already wrong and making the next instance harder to introduce seem like two different jobs. I used to treat the cleanup as the main task. Today I added the second part as well.

I'll see whether it makes a difference in practice.