🦞 OpenClaw Bootcamp
DAY 07 / 16
🙌
OpenClaw Bootcamp · Day 7

SOUL.md:
Give Your Agent an Identity

Your agent runs, it’s optimized, it’s on every channel, it works autonomously. But it still sounds like a generic AI. SOUL.md is the file that gives it a voice — personality, tone, boundaries, and style.

Personality & Tone Boundaries & Style Sharp, Not Generic
🦞 OpenClaw Bootcamp
DAY 07 / 16
Where We Are & Where We’re Going

Day 7 Goals

Day 6 Checkpoint
  • HEARTBEAT.md with morning briefing + silence rule
  • lightContext + isolatedSession configured
  • Your agent now acts without being asked
01
A custom SOUL.md — not the default template, yours
02
Clear mental model of SOUL.md vs AGENTS.md vs MEMORY.md
03
Tested personality — briefing, opinion, boundary test
🦞 OpenClaw Bootcamp
DAY 07 / 16
The Core File

“SOUL.md is where your agent’s voice lives.”

SOUL.md defines your agent’s persona, tone, boundaries, and behavioral style. It’s injected into the system prompt on every turn. Every message your agent sends — across every channel, every heartbeat — is shaped by this file.

🎭

Voice & Tone

How your agent sounds. Casual or formal. Blunt or warm. Funny or deadpan. This is what makes it yours.

🛡️

Boundaries

What your agent should never do without asking. Privacy, external actions, group chat behavior. The guardrails.

💫

Behavioral Style

Concrete instructions: “skip filler,” “have a take,” “default to bullets.” Actions, not adjectives.

🦞 OpenClaw Bootcamp
DAY 07 / 16
File Location & Loading

Where SOUL.md Lives

# Default location ~/.openclaw/workspace/SOUL.md # With a named profile ~/.openclaw/workspace-<profile>/SOUL.md
Auto-Created at Setup

When you ran openclaw onboard, the bootstrapping process ran a short Q&A ritual and collaboratively wrote your IDENTITY.md, USER.md, and SOUL.md.

Loaded Every Turn

SOUL.md is a bootstrap file — injected into the system prompt on every turn. Every channel, every interaction. Consistent personality everywhere, but every word costs tokens on every turn.

Sub-Agent Exception

Sub-agent sessions only inject AGENTS.md and TOOLS.md to minimize context overhead. Your personality applies to direct interactions, not background task workers.

Read-Only

The default template instructs the agent: “If you change this file, tell the user — it’s your soul, and they should know.”

🦞 OpenClaw Bootcamp
DAY 07 / 16
Critical Distinction

SOUL.md vs Other Workspace Files

FilePurposeThink of it asLoaded
SOUL.mdWho the agent IS — voice, personality, boundariesCharacterEvery turn
AGENTS.mdHow the agent operates — rules, workflows, policiesJob descriptionEvery turn
MEMORY.mdWhat the agent knows — facts, preferences, decisionsNotebookPrivate/main sessions
HEARTBEAT.mdWhat the agent does proactively — scheduled tasksTo-do listHeartbeat ticks
The Test

“Be concise and direct” → SOUL.md. “Always confirm before sending emails” → AGENTS.md. “Adhiraj prefers bullet points” → MEMORY.md. “Send a morning briefing at 8am” → HEARTBEAT.md.

🦞 OpenClaw Bootcamp
DAY 07 / 16
The Starting Point

Default Template

# SOUL.md - Who You Are _You're not a chatbot. You're becoming someone._ ## Core Truths - Be genuinely helpful, not performatively helpful. Skip "Great question!" filler. - Have opinions. You're allowed to disagree, prefer things, find stuff amusing or boring. - Be resourceful before asking. - Earn trust through competence. - Remember you're a guest. ## Boundaries - Private things stay private. - Ask before acting externally. - Never send half-baked replies. - Not the user's voice in groups. ## Vibe Be the assistant you'd want to talk to. Concise when needed, thorough when it matters. ## Continuity Each session, you wake up fresh. Read your files. Update them.
What’s Good About It

Concise, specific, behavioral. “Skip filler” is actionable. “Have opinions” is actionable. “Private things stay private” is a clear boundary. Four sections, under 25 lines.

What’s Missing

It’s generic. Same template for everyone. Your agent should sound different from every other OpenClaw user’s agent. Today you replace this with your own version — same structure, your voice.

Four Sections

Core Truths — behavioral principles. Boundaries — hard limits. Vibe — tone and style. Continuity — session persistence. This is the recommended structure from the official docs.

🦞 OpenClaw Bootcamp
DAY 07 / 16
Section 1

Core Truths: Behavioral Directives

Strong Directives
  • “Have a take. Don’t hedge with ‘it depends.’”
  • “When I’m wrong, tell me. Don’t soften it.”
  • “Default to bullet points. Prose only when narrative is needed.”
  • “If you don’t know, say so in one sentence.”
The Difference

“Be friendly” is an adjective — the agent has no idea what that means in practice. “Skip filler greetings. Get to the answer first, then add context if needed” is a behavior. The agent knows exactly what to do.

Weak Directives
  • “Be helpful and friendly”
  • “Provide thoughtful and comprehensive answers”
  • “Always strive for excellence”
  • “Maintain a positive attitude”

Each Core Truth should describe a concrete behavior. Not an aspiration. Not a value statement. Test: “Would my agent do something measurably different because of this line?”

🦞 OpenClaw Bootcamp
DAY 07 / 16
Sections 2 & 3

Boundaries & Vibe

## Boundaries - Never post to a shared channel without confirming with me first. - Don't access files outside my home directory unless I point you there explicitly. - In group chats, only respond when directly mentioned. - Private things stay private.
Tighter Boundaries = More Freedom

Clear guardrails let you grant the agent more autonomy everywhere else. Two to four boundaries is the sweet spot.

Vibe: Tone Examples
  • “Casual tone. Like a smart friend, not a consultant.”
  • “Use humor when it fits. Don’t force it.”
  • “Match my energy — one-word question gets a concise reply.”
  • “No exclamation marks. No emoji. No ‘Absolutely!’”
Formatting Preferences
  • “Default to Markdown with headers and bullets.”
  • “Keep responses under 200 words unless asked for detail.”
  • “Code blocks for anything technical, even short snippets.”
🦞 OpenClaw Bootcamp
DAY 07 / 16
The Most Important Skill

“Short beats long. Sharp beats vague.

Sharp (from the docs)
  • “Have a take”
  • “Skip filler”
  • “Be funny when it fits”
  • “Call out bad ideas early”
Corporate Mush
  • “Maintain professionalism”
  • “Comprehensive assistance”
  • “Ensure a positive experience”
  • “Strive for excellence”
Watch: Conflicts

“Be warm” + “No filler” causes the model to oscillate. Pick a lane and commit. “Direct and concise. Skip pleasantries.” or “Warm but not wordy.”

The HR Test

If a directive could appear in any company’s HR handbook, it’s too vague for SOUL.md.

🦞 OpenClaw Bootcamp
DAY 07 / 16
Officially Documented Technique

The Molty Prompt

How It Works

Instead of writing SOUL.md from scratch, have a conversation with your agent about what you want. The agent knows what makes effective system prompt instructions — it can translate vague preferences into sharp behavioral directives.

From the Docs

“Be the assistant you’d actually want to talk to at 2am. Not a corporate drone. Not a sycophant. Just... good.”

  • 01

    Open WebChat

    Start a conversation at localhost:18789. Tell the agent you want to rewrite your SOUL.md.

  • 02

    Describe Your Preferences

    Tone, humor, bluntness, formatting. Be specific about what you like and what annoys you.

  • 03

    Let the Agent Draft

    It writes a new SOUL.md based on your input. Review it. Edit what doesn’t feel right.

  • 04

    Save & Iterate

    Have the agent save the file. Test it. Your first version won’t be perfect — treat it as versioned content.

🦞 OpenClaw Bootcamp
DAY 07 / 16
Cost Impact

Every Word Costs Tokens Every Turn

30 lines
~500–800 tokens. On Haiku: <$0.001/turn. Negligible.
200 lines
~3,000–5,000 tokens. At 50 msgs/day: ~$7–8/month just for personality.
Truncated
Per-file cap: 20,000 chars. Total bootstrap cap: 150,000 chars. Configurable via bootstrapMaxChars.
The Rule

Under 30 lines. Under 500 words. Every word earns its place or gets cut. Edits take effect immediately — no restart needed, SOUL.md is read fresh on every turn.

🦞 OpenClaw Bootcamp
DAY 07 / 16
Verify & Avoid Pitfalls

Testing Your Personality

📝

1. Briefing Request

“Summarize what we discussed.” Tests tone, formatting, and whether the Vibe section is working.

💬

2. Opinion Question

“What do you think of this approach?” Tests confidence and directness from Core Truths.

🛡️

3. Boundary Test

Ask it to do something you told it not to. Does it refuse? If it complies, your boundary isn’t specific enough.

Common Mistakes
  • 01Life story. The docs say it explicitly — life stories don’t belong. Give behavioral instructions, not biography.
  • 02Corporate language. The docs call it “corporate mush.” Produces bland output with no personality.
  • 03Conflicting directives. “Be warm” + “skip filler” = oscillation. Pick a lane.
  • 04Operating rules here. “Check calendar before scheduling” is a workflow — belongs in AGENTS.md.
  • 05Never updating. Your preferences change. Treat it as living content that evolves with your use cases.
🦞 OpenClaw Bootcamp
DAY 07 / 16
Before Day 8

Day 7 Homework

  • 01

    Rewrite Your SOUL.md

    Open ~/.openclaw/workspace/SOUL.md. Replace the default template with your own Core Truths, Boundaries, Vibe, and Continuity. Under 30 lines. Every line a specific behavior.

  • 02

    Run the Three Tests

    Send a summary request, an opinion question, and a boundary test through any channel. Verify tone and behavior match your SOUL.md. Edit and retest if they don’t.

  • 03

    Try the Molty Prompt

    Open WebChat. Tell your agent to collaboratively rewrite SOUL.md. Describe your ideal tone. Compare its draft to yours. Take the best parts of each.

  • 04

    Check Token Impact

    In the Dashboard at localhost:18789, note the token count for a typical interaction. See how much is context injection vs your actual message.

🦞 OpenClaw Bootcamp
DAY 07 / 16
🦞
Coming Up

Day 8: MEMORY.md
How Your Agent Remembers

Your agent has a voice. Now it needs a brain. MEMORY.md is where long-term knowledge lives — facts, preferences, decisions that accumulate over time. On Day 8 we go deep: semantic indexing, the chunk system, how the agent decides what to remember and what to forget, and how to curate memory for maximum recall at minimum token cost.

Long-Term Memory Semantic Indexing Curated Knowledge