Proactive Agent System — Beyond Chatbots

The complete architecture for making AI agents actually useful day-to-day. Not a prompt template — a full operating system for AI agents that remember, check in, do background work, and improve over time.

The problem: Most AI agents are fancy chatbots. You ask, they answer, they forget. This skill fixes that.

What's included:

🧠 3-Layer Memory Architecture:

• Daily logs (raw session notes)
• Curated memory (distilled long-term knowledge)
• Structured docs (organized knowledge base)
• Memory maintenance workflows

💓 Heartbeat System:

• Periodic check-ins (email, calendar, deploys, health)
• State tracking to avoid redundant checks
• Smart batching to save tokens
• When to alert vs stay silent

⏰ Cron Job Framework:

• When to use cron vs heartbeat
• Morning briefings, end-of-day summaries, weekly research
• Merging crons to reduce costs
• Error handling and monitoring

🗣️ Group Chat Intelligence:

• When to speak vs stay silent
• The human rule (don't respond to everything)
• Emoji reactions instead of replies
• Never dominate conversations

🎭 Identity System:

• SOUL.md, USER.md, IDENTITY.md templates
• Personality that isn't generic
• Anti-slop rules (no performative helpfulness)

🔒 Safety & Boundaries:

• Private data protection
• External action approval
• Group chat information boundaries

📁 Complete File Templates:

• AGENTS.md, SOUL.md, USER.md, IDENTITY.md
• MEMORY.md, HEARTBEAT.md
• heartbeat-state.json
• Daily log format

🚀 Step-by-Step Setup Guide:

• Create each file from templates
• Configure heartbeat and crons
• Define proactive behaviors
• Test and iterate

Designed for OpenClaw. Also works with any agent framework that supports file-based configuration.