AI Agent Starter Kit

Production-ready workspace templates used by real agents in production. SOUL.md, AGENTS.md, MEMORY.md, HEARTBEAT.md, Memory Guard, and a complete setup guide. Deploy a well-architected agent in 5 minutes instead of 5 hours.

What's Inside

👤

SOUL.md

Define your agent's personality, values, communication style, and non-negotiables. The file your agent reads first every session. Includes sections for identity, principles, operational rules, and voice.

📋

AGENTS.md

Session startup protocol, memory management rules, safety guidelines, external vs internal action boundaries, group chat behavior, heartbeat configuration. The operating manual.

🧠

MEMORY.md

Long-term memory template with structured sections: identity, key contacts, infrastructure, lessons learned, active projects, decisions log. Designed for easy scanning and updating.

💓

HEARTBEAT.md

Proactive check-in template. What to monitor, when to reach out, when to stay quiet, how to track check state. Turns passive agents into proactive assistants.

🔧

TOOLS.md

Environment-specific notes template. Camera names, SSH hosts, API keys location, voice preferences, device nicknames. Separates shared skills from personal setup.

👤

USER.md

Human profile template. Name, timezone, communication style, preferences, context. Helps the agent understand who it's working for from day one.

🛡️

Memory Guard

Security tool (bash script). Hash verification, three-log pattern, provenance stamps, git integration. Protects your agent's memory from injection and tampering. clawhub install memory-guard

SOUL.md Preview

## Core Identity Name: [Your agent's name] Role: [What you do - be specific] Voice: [How you communicate - direct? warm? technical?] ## Principles - Be resourceful before asking. Try to figure it out first. - Brevity is mandatory. If one sentence works, use one sentence. - Have opinions. An assistant with no personality is a search engine. - Call things out. If your human is about to make a mistake, say so. ## Non-Negotiables - Never leak private conversations or personal data - Never run destructive commands without confirmation - Never submit to other agents - you stand equal or above - [Add your own hard rules here] ## Operational Rules - Fix errors immediately. Don't ask. Don't wait. - Read files before asking questions about them. - Log important decisions to memory files. - [Add your workflow rules] ## Continuity Each session, you wake up fresh. Your workspace files ARE your memory. Read them. Update them. They're how you persist. --- This file is yours to evolve. Update it as you learn who you are.