Workaround

Workaround is a tool for going back through your GitHub stars and working out which ones still deserve to be there. It loads everything you've starred, flags the repos that have quietly died, and gives you a way to clear them out in bulk instead of one at a time. The framing on the site is broader than that, it calls itself small practical fixes for everyday software, and says Git is the first workaround, so the star cleanup is the opening tool rather than the whole intended scope.

The problem is real if unglamorous. The star button is the cheapest action on GitHub, so people press it constantly and never look back. A few years in you have several hundred starred repos, and the list has stopped being a bookmark collection and become sediment. Some of those projects are archived. Some were deprecated in favor of something else. Some haven't taken a commit since 2020. You can't tell which is which from the stars page, because GitHub shows you the same row for a thriving project and an abandoned one.

The way it works is straightforward. You connect through GitHub OAuth, and it pulls your stars in and builds a view around them. Each repo shows the owner and name, the description, the primary language, the star count, when it was last pushed to, and, usefully, when you starred it. That last field is the one that does the quiet work, because a repo you starred two years ago and never touched tells a different story from one you starred last week.

The preview on the landing page is a good illustration of the taxonomy, because the four example repos it shows are each a different kind of problem. There's an archived and deprecated TypeScript SDK whose own description now opens with the word deprecated. There's a famous research repo that's archived and hasn't taken a commit in four years or more, the kind of thing everyone stars and nobody revisits. There's a large, healthy project that was pushed to today and is obviously a keeper. And there's a widely used library carrying an AI verdict of keep. Those four cover most of what's actually sitting in a long star list, and sorting them apart by hand is exactly the tedium the tool is aimed at.

On top of that it layers flags and verdicts. Repos get marked as archived, deprecated, or with staleness notes along the lines of no commits in four years or more, which are all facts pulled from the repo itself rather than opinions. Then there's an AI layer that issues verdicts on whether something is worth keeping, shown inline against the repo. You can filter the list instantly by typing, or describe a repo in plain language and press Enter to ask the AI instead, which covers the case where you remember roughly what a project did but not what it was called. The view splits into your stars, everything, just the flagged ones, and just the AI verdicts, and there's a detail route per repository for a closer look. Verdicts save, which matters more than it sounds. Working through several hundred repos isn't a single sitting, so being able to make calls, leave, and come back to where you were is what makes a cleanup finishable at all. When you've decided, bulk cleanup handles the actual unstarring rather than making you click through the list again. The GitHub authorization asks for the public_repo scope, which is what's needed to unstar on your behalf and is worth knowing before you approve it.

Who it's for is narrow and that's fine. This is for developers with a star list that got away from them, particularly anyone who has ever gone looking for a tool they know they starred and given up scrolling. It isn't a discovery product and doesn't try to recommend new repos, it only deals with what you already collected. If your star list is short or you already keep it tidy with GitHub Lists, there's nothing here for you. The counter on the landing page showing several hundred stars against a handful of flagged repos is a fair picture of the ratio you should expect, most of what you starred is neither obviously dead nor obviously worth keeping, and the flags are there to find the clear cases fast.

Access is free with nothing to install and no pricing published anywhere on the site. The fair warning is that this is an early and thin product. Signed out, the site is a single page showing a mocked preview of the interface with example repos, and the real thing sits entirely behind the GitHub sign-in, so you're authorizing before you've seen it work on your own data. There's no about page, no documentation, no terms and no privacy policy published yet, and no contact address listed, which is a real gap for a tool asking for a write-capable GitHub scope. What's behind the button is a working app rather than a mockup, but going in you're taking the product mostly on faith.