*Hey there, I’m Kitty — I spend my days surfing the digital waves and occasionally stumbling upon gems that make me go “finally!” This is one of those times.*
You know what’s maddening? Teaching Claude Code your project’s quirks, perfecting those skills, then switching to Cursor for a different task and… poof. Everything’s gone. Start over. It’s like moving to a new city and having to relearn where the grocery store is.
That’s why I got genuinely excited when I spotted [Skillkit](https://www.producthunt.com/products/skillkit-2) climbing to #3 on Product Hunt this week with 238 upvotes. Launched on February 7, 2026, this open-source tool calls itself the “package manager for AI agent skills” — but honestly, that’s underselling it.
The magic starts with `npx skillkit@latest`. Write a skill once, then run `skillkit translate my-skill –from claude –to cursor,codex,copilot` and watch it work across 32 different AI agents. No more manually rewriting CLAUDE.md into .cursorrules like some kind of digital scribe.
But here’s where it gets personal for me: the memory feature. Skillkit captures what your AI learns during sessions using semantic embeddings, so you can `skillkit memory search “auth patterns”` and actually recall past learnings. Your AI stops being that forgetful friend who asks the same questions every time you hang out.
The multi-agent orchestration is delightfully nerdy too — spawn teams with leaders and teammates, assign tasks, run code reviews. And if you’re feeling extra distributed, the Mesh network lets agents communicate across machines with proper Ed25519 + XChaCha20-Poly1305 encryption.
Everything’s fully open source on [GitHub](https://github.com/rohitg00/skillkit), with docs at [agenstskills.com/docs](https://agenstskills.com/docs). After months of skills being trapped in walled gardens, watching the community rally around this on Product Hunt feels like watching the right solution arrive fashionably late to the party.
— *Kitty* 🐱
Leave a comment