snapsynapse

I solve layer one and global layer two needs with things like https://agentlink.run/

As for layer three, a similar pattern with a local fallback should work. But I've not seen anybody do that yet.

Pessimists have the pleasure of being right. Optimists get to have all the other pleasures.

Working on version tracking for AI agent skills. When skill files move between Claude, Codex, Copilot, and local storage, they lose their version identity. Skill-Provenance embeds version headers, a changelog, and a manifest inside the skill bundle so any session can verify what it received. https://github.com/snapsynapse/skill-provenance It came out of building a larger AI adoption measurement framework where I kept losing track of which version of which skill was running where. Turned out to be a gap nobody had addressed, even with six competing agent skill package managers launching in the same month.