I've been using OpenClaw daily and I love it. One thing I noticed — frameworks like Hermes Agent (47K stars, Nous Research) and SkillClaw (AMAP-ML / 高德地图) have this "self-learning" feature where the agent automatically learns from conversations and writes reusable skills in the background, so it gets smarter the more you use it.

OpenClaw already has the best ecosystem (Telegram, WebChat, Discord, cron, etc.), but it still relies on users writing skills manually. If OpenClaw could learn from real interactions the way Hermes and SkillClaw do, it would be incredible.

I'm not a developer so I can't propose a specific architecture. Just suggesting the team takes a look at how those projects handle self-evolution — the core ideas are worth borrowing. "If you can't beat them, learn from them."

Thanks for building such a great tool!