The way I star GitHub repositories is pretty much the same as how I like and collect videos on bilibili: I just click them on a whim. Whenever I see a project that looks even remotely interesting, my brain always goes “might be useful someday” or “I’ll take a look later” — and my finger clicks the button before my brain finishes the thought.
That’s how, over the years, my star count quietly climbed past 800. To be honest, I have no idea what those 800+ projects actually are — most of them were starred because they seemed useful back then, and then never opened again.
I’d thought about cleaning up before, but it was too much hassle
It’s not that I never wanted to clean up. I’ve tried some star-management tools before — the kind that pull out your star list and group it by tag. But they were still a pain: after the sorting, I still had to look at every single one and click through it by hand. Going through hundreds of stars one manual operation at a time would wear my finger out. So the cleanup kept getting postponed, all the way until today.
With an agent, everything simplifies
This time I simply delegated the job to an agent on Multica: pull out all my stars, match them against my existing 14 star lists, suggest tags for the untagged ones, and finally produce a complete table.
The agent’s workflow was clear:
- Pulled all 808 of my stars via the GitHub API (
gh api) - Matched them against my 14 existing star lists (AI Or ML / Self-Host / Bioinformatics / AppDev / DataScience / We-media / etc.)
- Kept the 309 already-tagged stars as-is, then judged the remaining 499 untagged projects one by one by their descriptions, suggesting tags that all fell into the existing 14 categories
- Output everything as a markdown table for me to review
After I got the table, my part was simple: keep what I want, delete what I don’t. Then I told the agent to batch-unstar everything not on the list.
Results:
- 335 stars were unstarred in total, bringing the count from 808 down to 473, exactly matching my keep-list
- For the kept projects, the 174 that needed tagging — 274 tag entries in total — were also written back to the corresponding star lists by the agent in one go
The operations I needed to do were very few: open codium, review the table, delete the projects I didn’t want. Everything else — fetching, matching, batch-unstarring, writing back tags — was done by the agent.
Some thoughts
Those dedicated star-management tools are probably going to lose a chunk of their market. The core problem they solve — organizing scattered stars into categories and helping me bulk-clean them — a general-purpose agent can now do better. And the agent’s advantage isn’t just automation: it can intelligently judge which category a project belongs to based on its description, call the GitHub API directly for batch operations, and even caught a subtle detail for me along the way (that writing back to star lists requires the GraphQL API plus the user scope).
I’ll keep clicking stars on a whim, of course — but at least the cleanup part won’t give me a headache anymore.
