Clean Up Your GitHub Account

At some point my GitHub account turned into the digital equivalent of a garage full of broken chairs, mystery cables, and boxes labelled "might be useful later".

I had created well over 100 repositories. Some were real projects. Some were experiments. Some were "I just want to test one tiny thing" repos that got a single commit and were then quietly abandoned to the elements.

After looking at the pile for long enough, I built github-smash. It is a small tool that runs through a GitHub Actions workflow and deletes every repository that is not on a whitelist. Which is a mildly alarming sentence, but also very satisfying.

Why I Built It

I did not want to manually inspect repo after repo asking:

• Is this still useful?
• Does this contain anything important?
• Why did I name this one test-final-2-actual?
• Why are there four versions of the same half-finished idea?

That kind of cleanup sounds manageable until you are 37 repositories deep and opening another abandoned prototype from a random Tuesday night.

So instead of deleting things one by one, I wanted something boring and repeatable:

1. Keep the projects I know I care about.
2. Preview what is about to disappear.
3. Remove the rest.

How github-smash Works

The setup is intentionally simple:

• A repo-whitelist.yml file contains the repositories to keep in owner/repo format.
• A GitHub Actions workflow fetches all repositories owned by the authenticated user.
• The workflow compares that full list against the whitelist.
• Anything not on the whitelist is either reported in dry_run mode or deleted for real.

A tiny whitelist looks like this:

repositories:
  - allistera/allistera.github.io
  - allistera/github-smash
  - allistera/things-sorter

The workflow uses a token stored as a repository secret, so the cleanup logic lives in GitHub and can be triggered from the Actions tab without needing to run a local script.

Why I Chose a Whitelist Instead of a Blacklist

A blacklist sounds safer at first. You make a list of repos to delete and leave everything else alone. Very sensible. Very measured. Very adult.

The problem is that my account was already far beyond that point.

I did not have a neat list of five obvious repositories to remove. I had a sprawling heap of forgotten side projects, throwaway tests, abandoned demos, and "I will sort this out later" repos. A blacklist would have meant identifying every single repo I was confident could be deleted. That was exactly the job I was trying to avoid.

The whitelist approach flipped the decision:

• If I know I want to keep it, it goes on the whitelist.
• If I am unsure, the dry run gives me a chance to catch it before anything happens.
• If it is not important enough to make the keep list, it is probably part of the mess.

It also matched the risk profile better. There were too many old repositories where I could not be completely sure what was in them anymore. Old config. Forgotten experiments. Possibly some ancient secret-shaped mistake from a less disciplined era. I did not want to pretend I had perfect certainty about what was safe to throw away.

The whitelist forced me to decide what deserved to survive, instead of trying to maintain a giant list of everything I thought was probably junk.

Dry Run First, Chaos Second

github-smash has a dry_run mode, which is the feature that stops this whole idea becoming performance art.

In dry run, the workflow tells you:

• how many repositories it found
• how many are on the whitelist
• what would be deleted

That gives you one last chance to spot the repo you forgot about before the workflow gets enthusiastic.

Only when the output looks right do you run it again with dry_run set to false.

A Slightly Unhinged Way to Tidy Up

This is not enterprise-grade lifecycle management. This is just me looking at a GitHub account full of neglected projects and deciding it was time to stop pretending I would come back to weekend-tooling-test one day.

If your GitHub account has also become a small software graveyard, github-smash might help. Add the repos you actually care about, run the workflow in dry-run mode, and let the cleanup begin.