These are the skills I run my own agents on, every day. Now I am opening them.

A skill is instructions your agents carry out with your machine's access. That is real power, and most skill packs ask you to take it on faith. I would rather not. So every skill here is two files: a SKILL.md you read in plain language, and a protocol.yon you validate yourself against the public YON (YounndAI Object Notation) parser, before you trust it to run. Fifty skills today. Thirty-six carry a protocol you can check. Fourteen are the readable doc while their protocols land.

I did not write these to publish them. I wrote them because I needed them: review, planning, priming, deliberation, orientation. I still run them on real work, and I am still sharpening them, so more will come and the ones here will keep changing. Opening them was the natural step. The name says the intent.

They run on YON, the same grammar under everything else I build. A skill's rules and gates should be data you can check, not prose you hope a model follows.

One more thing worth saying plainly. These run with your agents' machine access, the same as any skill you install. So the two files are the contract: read the SKILL.md, validate the protocol.yon, then decide. Trust is earned by reading, not extended on faith.

This is an independent allemaar project. Apache-2.0, free to fork and adapt, DCO sign-off on contributions.

Knowledge is free. What you do with it is the part that matters.

Read the skills. Run the ones you trust.

I Walk the Way.

Repo: github.com/allemaar/open-skills · Install: a Claude Code plugin, or clone it.

open-skills is an independent allemaar project. YON and YounndAI are trademarks of MARLINK TRADING SRL.