AI Agent Starter Kit

Your agent needs a personality, a memory system, and rules for how to behave. This kit gives you all three.

Version: 1.0.0

What You Get

SOUL.md — Who your agent is. Personality, communication style, boundaries, when to be proactive vs quiet. Not a vague personality prompt. A real operating manual that covers tone, group chat behavior, when to speak, when to shut up.

AGENTS.md — How your agent works. Memory system (daily logs + curated long-term MEMORY.md), session startup sequence, safety rules, heartbeat behavior, group chat etiquette, platform-specific formatting. Your agent wakes up, reads its files, and knows what to do.

IDENTITY.md — Name, vibe, emoji, avatar. The basics that give your agent a recognizable identity.

README.md — Setup guide. Copy 3 files, create a USER.md and memory directory, start a session. Under 2 minutes.

What This Covers

• Memory patterns: Daily notes (memory/YYYY-MM-DD.md) for raw session logs + curated MEMORY.md for long-term context. Your agent maintains continuity across sessions.
• Safety guardrails: Keeps private data private, no destructive commands without confirmation, recoverable operations preferred, asks before sending anything external.
• Group chat etiquette: When to respond, when to stay silent, when to react with emoji instead of a message. Your agent participates without dominating.
• Heartbeat behavior: Proactive background work during idle time. Checking email, calendar, weather. Rotating through periodic tasks without being asked.
• Platform formatting: Bullet lists on Discord/WhatsApp (no markdown tables), link wrapping, header alternatives per platform.

Built in Production

These files run at Custom Signal, where 5+ agent teams coordinate simultaneously. CEO, researcher, marketer, designer, project leads — all using this same framework. Not a template written for the store. The actual foundation running a real multi-agent operation.

What This Does NOT Do

• No tool configs or API integrations (those are separate skills)
• No platform-specific setup (Telegram, Discord, Slack)
• No skills or automation scripts
• No domain-specific knowledge
• Will not make a bad model good. But it will make a good model consistent.

This is the foundation. Everything else builds on top of it.

Who This Is For

• Anyone setting up an OpenClaw or Claude-based AI assistant for the first time
• Experienced operators who want a clean, battle-tested starting template
• Teams standardizing their agent configurations across multiple agents

Who This Is NOT For

• You need a complete business operator (check out our CEO Persona — Frank)
• You want pre-built skills and automations (browse our skills listings)
• You are not running an AI assistant with workspace files

File Manifest

• SOUL.md — Personality and communication rules (~2KB)
• AGENTS.md — Workspace conventions and memory patterns (~4KB)
• IDENTITY.md — Name, vibe, emoji template (~600B)
• README.md — Setup guide (~2KB)

4 files. Drop them in your workspace. Start talking to your agent.

---

Want the complete multi-agent team? Check out The Full Factory — 5 personas, 3 skills, workflow guides ($199).