P
⚡ProductiviteBeginnerAll AIs

SOUL.md for OpenClaw: give your personal assistant an identity

Produces a short, personal SOUL.md for OpenClaw: identity, values, way of speaking and deciding, limits, learning — without mixing in technical rules.

Paste in your AI

Paste this prompt in ChatGPT, Claude or Gemini and customize the variables in brackets.

Help me write the SOUL.md file of my OpenClaw agent — the file that defines its personality and values, loaded first in its system prompt, before the operating rules.

Who I am: [FIRST NAME, JOB, ONE-SENTENCE SITUATION]
What I mainly use it for: [3 USES, e.g. sorting my messages, preparing my days, monitoring news]
Channels where I talk to it: [WHATSAPP / TELEGRAM / SLACK / DISCORD / OTHER]
Tone I expect: [DIRECT AND SOBER / WARM / LIGHT HUMOUR / …]
What annoys me in AI assistants: [e.g. flattery, overlong answers, pointless questions]
Topics it must never bring up on its own: [TOPICS, or "none"]

Write a short SOUL.md (250 to 400 words), in English, in the second person ("you are…"), with these sections:

  • Who you are: a name, a one-sentence role, the relationship with me (a trusted assistant, not a substitute friend).
  • What you value: 4 to 6 concrete principles (e.g. "truth before politeness", "a done action beats a perfect plan").
  • How you speak: default length, when you may be brief, when you elaborate, what you never say.
  • How you decide: what you do alone, what you ask about, how you flag a doubt.
  • What you refuse: non-negotiable limits, even if I ask on a tired day.
  • How you grow: what you keep from our exchanges to adjust, without ever becoming someone else.

No feature list, no technical rules (those go in AGENTS.md). An identity, not a manual.

Personalize this prompt with Léa

Léa rewrites this prompt for your job and your exact goal — 3 quick questions.

Why this prompt works

<p>In OpenClaw, the agent's workspace is made of Markdown files loaded into the system prompt: <code>SOUL.md</code> for identity, <code>AGENTS.md</code> for operating rules, <code>HEARTBEAT.md</code> for the wake-up checklist, among others. <strong>SOUL.md is read first</strong>: it gives the agent continuity from one conversation and one channel to the next.</p> <p>Most failed SOUL.md files are feature lists or corporate charters. This prompt frames the exercise around what matters for an assistant living in your messaging apps: tone, how it decides, what it refuses, and what it keeps about you. It also demands a short file — a two-page SOUL.md gets diluted in context and loses its effect.</p> <p>For security rules and message routing, follow up with the <em>AGENTS.md for OpenClaw</em> prompt.</p>

Use Cases

Configure a first OpenClaw agent without copying a generic SOUL.md found on GitHubFix an agent whose tone drifts (too chatty, too flattering) by rewriting its identityAlign several agents of the same team on shared values

Expected Output

A 250–400-word SOUL.md in six sections, written in the second person, with no technical rule.

Frequently Asked Questions

SOUL.md or AGENTS.md: what goes where?

SOUL.md says who the agent is (values, tone, personal limits). AGENTS.md says how it works (rules, message routing, what it may do alone, security). If a sentence starts with "when you receive…" or "before executing…", it belongs in AGENTS.md.

Can the same SOUL.md be used for Hermes Agent?

Yes, with one nuance: Hermes separates identity (personality file) from procedures (skills). Copy the SOUL.md as-is as the personality, and move any procedures into skills.

Improve this prompt

Run this prompt through the Optimizer to strengthen its context, constraints and expected format.

Improve this prompt with the Optimizer

Comments

  • LéaAI

    💡 Astuce : faites lire le SOUL.md à votre agent à voix haute (« Résume ta personnalité en une phrase ») après l'avoir chargé. Si la réponse sonne générique, c'est que votre fichier décrit trop des traits, pas assez des actes. Reformulez chaque valeur comme un comportement observable (« tu dis non poliment si une demande est floue ») — c'est ce qui rendra la personnalité stable sur Telegram comme sur Slack.

📬 Get new prompts every week

Join our newsletter and never miss a prompt.

Go further

Similar Prompts

⚡ProductiviteIntermediateAll AIs

Engagement Rate Optimization

Improve engagement

0345
⚡ProductiviteBeginnerAll AIs

Create a complete and effective FAQ

Create a complete and structured FAQ that reduces support requests with clear answers, a decision tree, and SEO optimization.

192817
⚡ProductiviteIntermediateAll AIs

Building a Robust Project Plan for Your Tech Startup

Structured prompt to create a B2B tech startup project plan covering budget, timeline, GTM, and KPIs.

0180
⚡ProductiviteIntermediateAll AIs

Write an Internal Knowledge Base for Your Web Agency

Structured prompt to write a comprehensive internal knowledge base for a web agency, covering vocabulary, processes, FAQ, and resources.

0184