Most AI agents sit idle until you tell them what to do. This one watches for what you might need next, surfaces ideas you didn't think to ask for, and gets smarter over time. The memory system flushes context to disk before the window fills (so your agent doesn't wake up amnesiac after a long conversation), reverse prompting suggests work that would help without being pushy, and self-healing patterns let it diagnose and fix its own issues before bothering you. Security hardening keeps it from following instructions buried in external content. Use this if you want an agent that compounds knowledge across sessions and creates value without waiting to be asked.
npx skills add https://github.com/sundial-org/awesome-openclaw-skills --skill proactive-agent