Hey all,
Over the past year I noticed I was maintaining the same AI workflows in multiple AI tools such as Claude code and Codex.
Whenever I improved one workflow, I had to remember everywhere else I'd copied it and most of the time I didn't... so I quickly went the "dotfiles" route where I symlinked everything together. Then I thought, this works for me since I'm used to my CLI but others might not have such a good time.
Had a few talks with people, had some discussions, and in the end I ended up building Skillful.
Instead of treating prompts or skills as something that lives inside a single AI tool, Skillful treats them as ordinary folders on disk. A skill consists of a SKILL.md or AGENT.md together with whatever else it needs, examples, scripts, screenshots, documentation, etc.
The application lets you:
- manage everything from one local library - install the same workflow into multiple AI tools using links instead of copies - Sync your library to multiple computers with Github - import public GitHub repositories containing skills - detect and repair broken installs
Everything stays on YOUR machine. There are NO accounts, NO telemetry, and NO backend. It's essentially a GUI around a filesystem with symlinks, because I wanted the files to remain useful even if the application disappeared one day.
I also started an "awesome" repository with reusable skill packs that anyone can publish or contribute to:
https://github.com/Mastermindzh/awesome-skillful I'd love feedback on a few things:
Does the "one source, many AI tools" approach make sense for you too? Would you rather manage these as plain Git repositories? What AI tools should be supported next? If you already use reusable skills, how are you organising them today?
Looking forward to hearing what you think!
PS: I know that Vercel Labs also built npx skills whilst I was building Skillful, I think of this as an alternative :P
Comments URL: [https://news.ycombinator.com/item?id=48932688](https://news.ycombinator.com/item?id=48932688)
Points: 1