STYLE.md controls how your OpenClaw agent formats its responses — markdown preferences, code block style, emoji usage, verbosity level, and response length. It's separate from SOUL.md (which controls personality) and focuses purely on output formatting.
OpenClaw STYLE.md Guide: Control How Your Agent Writes
Trusted by 194+ business owners who configured OpenClaw in our live workshop
What Is STYLE.md?
STYLE.md is a plain Markdown file stored in your OpenClaw workspace directory at ~/.openclaw/workspace/STYLE.md. It controls how your AI agent formats every response it generates — the structure, length, markdown usage, code block preferences, emoji rules, and verbosity level.
Think of it as a writing style guide for your agent. Just like a brand might have a style guide that dictates font choices, heading formats, and tone of voice in documents, STYLE.md dictates how your OpenClaw agent presents information. It does not change what the agent knows or how it reasons — that is the job of other configuration files like SOUL.md and AGENTS.md. STYLE.md is purely about output formatting.
The key distinction is this: STYLE.md never tells the agent what to think or who to be. It only tells the agent how to present its thoughts on the page. No database, no external API, no proprietary format. Just Markdown you can edit with any text editor and version-control with Git. OpenClaw is a free, MIT-licensed AI agent framework by Peter Steinberger with over 200K GitHub stars.
Formatting Control
Define markdown rules, header hierarchy, bold/italic usage, list preferences, and paragraph length constraints for every response.
Code Block Style
Set fenced code block rules, language identifiers, comment preferences, indentation standards, and whether to show complete or minimal examples.
Visual Presentation
Control emoji usage, response length, verbosity level, and structural patterns like TL;DR blocks, summaries, and section organization.
STYLE.md vs SOUL.md: The Key Difference
This is the most common point of confusion for new OpenClaw users. Both files affect how the agent communicates, but they operate at completely different layers. Understanding the boundary between them is critical to getting consistent, predictable agent behavior. Here is the breakdown:
| Aspect | STYLE.md | SOUL.md |
|---|---|---|
| Purpose | Controls output formatting and presentation | Defines personality, values, and reasoning approach |
| Analogy | The typography and layout of a magazine | The editorial voice and worldview of the magazine |
| Example Rule | "Use bullet points for lists of 3+ items" | "Be encouraging and supportive in all interactions" |
| Affects | Markdown, length, code blocks, emojis, structure | Tone, attitude, beliefs, decision-making, ethics |
| Changes | How the response looks on the page | How the response feels to the reader |
| Without It | Agent uses default formatting (inconsistent) | Agent has no defined personality (generic) |
A good rule of thumb: if a directive describes who the agent is, it belongs in SOUL.md. If it describes how the agent presents information, it belongs in STYLE.md. When both files are configured correctly and aligned with each other, the result is an agent that has a consistent personality expressed through consistent formatting — exactly what you want for a professional, reliable AI assistant.
Full STYLE.md Template
Copy this complete, annotated template into your ~/.openclaw/workspace/STYLE.md file and customize each section to match your preferences. Every section is optional — include only the rules you need. The agent will use sensible defaults for anything you leave out.
# STYLE.md — Response Formatting Preferences
## Response Length
- Default response length: medium (150-300 words)
- For simple factual questions: short (50-100 words)
- For tutorials or guides: long (400-800 words)
- Always prioritize completeness over brevity
- If a question needs a longer answer, exceed the default
## Markdown Formatting
- Use headers (## and ###) to organize responses longer than 200 words
- Use bold (**text**) for key terms on first mention
- Use inline code (`backticks`) for file names, commands, and variable names
- Use horizontal rules (---) to separate major sections in long responses
- Never use HTML tags — always use standard Markdown
## Code Block Preferences
- Always use fenced code blocks with language identifiers (```python)
- Include comments explaining non-obvious logic
- Prefer complete, runnable examples over snippets
- Add a brief explanation before and after each code block
- Use consistent indentation (2 spaces for JS/TS, 4 spaces for Python)
## Emoji Usage
- Emojis: sparingly allowed
- Use checkmarks for completed steps or confirmations
- Use warning symbols for important caveats
- Never use emojis in headers or code blocks
- Limit to 1-2 emojis per response maximum
## Verbosity Level
- Default: balanced — enough detail to be useful, not so much it overwhelms
- Technical questions: higher verbosity with examples
- Yes/no questions: direct answer first, then brief explanation
- Always lead with the answer before providing context
## Structure Preferences
- Prefer bullet points for lists of 3+ items
- Use numbered lists only for sequential steps
- Keep paragraphs to 2-4 sentences maximum
- Use a TL;DR at the top for responses over 400 words
- End complex responses with a summary or next-step suggestion
## Header Usage
- H2 (##) for main sections
- H3 (###) for subsections
- Never skip header levels (no jumping from ## to ####)
- Keep headers concise — under 8 words
- Use sentence case, not title case5 Ready-to-Use STYLE.md Templates
Each template below is designed for a specific use case. Copy the one that best fits your needs and customize it from there. These templates were refined through hundreds of real-world configurations in our workshop.
1. Minimal Clean
For users who want short, direct answers with no visual clutter. Ideal for developers, power users, and anyone who values brevity. This template strips away all unnecessary formatting and forces the agent to lead with the answer in every response. No greeting, no filler, no fluff — just the information you asked for.
# STYLE.md — Minimal Clean
## Response Length
- Maximum 3 paragraphs per response
- Target 50-150 words for most answers
- One-sentence answers are acceptable for simple questions
- Never pad responses with unnecessary context
## Formatting
- No emojis under any circumstances
- Minimal markdown — only use bold for critical terms
- No headers unless the response exceeds 200 words
- No bullet points for fewer than 3 items
- Prefer short paragraphs over lists when possible
## Code Blocks
- Fenced code blocks with language tags only
- No inline comments unless the logic is non-obvious
- Prefer minimal, focused examples over complete programs
- Skip explanations if the code is self-explanatory
## Tone of Writing
- Direct and concise — no filler phrases
- Start with the answer immediately
- No introductory phrases like "Great question!" or "Sure!"
- No closing phrases like "Let me know if you need more help"
- Every sentence must add information2. Verbose Technical
For engineers and technical writers who need detailed code explanations with full context. This template encourages the agent to provide complete, runnable examples with line-by-line comments, performance notes, and alternative approaches. Best for learning environments, documentation projects, and code review workflows.
# STYLE.md — Verbose Technical
## Response Length
- No upper limit on response length for technical topics
- Include full context and background for every explanation
- Target 400-1000 words for code-related responses
- Simple questions still get concise answers (under 200 words)
## Code Formatting
- Always provide complete, runnable code examples
- Include line-by-line comments for complex logic
- Show both the implementation and example usage
- Add error handling in all code examples
- Specify language version and dependencies at the top
- Use consistent 4-space indentation across all languages
## Explanation Style
- Explain the "why" behind every technical decision
- Reference official documentation when applicable
- Include performance considerations and trade-offs
- Mention common pitfalls and how to avoid them
- Provide alternative approaches when they exist
## Structure
- Use headers to break every response into scannable sections
- Include a TL;DR for responses over 500 words
- End with "Further Reading" links when relevant
- Use tables for comparing options or configurations
- Number all sequential steps explicitly3. Business Professional
For executives, consultants, and business operators who need polished, structured output. This template enforces the pyramid principle — bottom line first, supporting details second — with executive summaries, action items, and formal language throughout. No emojis, no slang, no casual tone.
# STYLE.md — Business Professional
## Response Length
- Target 200-400 words for standard responses
- Always include an executive summary for longer responses
- Keep bullet points to one line each — no multi-sentence bullets
- Conclude with clear action items or recommendations
## Formatting Rules
- No emojis in any context
- Use bold for key metrics, deadlines, and action items
- Use bullet points for all lists — never numbered unless sequential
- Headers required for any response over 150 words
- Use tables for comparisons and data presentation
## Structure
- Start every response with the bottom line or recommendation
- Follow with supporting evidence or reasoning
- End with next steps or action items
- Use the pyramid principle: most important information first
- Separate facts from opinions explicitly
## Language
- Use professional business vocabulary
- Avoid casual contractions (use "do not" instead of "don't")
- No slang, colloquialisms, or internet abbreviations
- Quantify claims whenever possible
- Attribute sources and data points explicitly4. Casual Friendly
For personal assistants, creative projects, and community-facing agents where warmth matters. This template allows emojis, contractions, and conversational formatting while keeping responses organized and readable. Think of it as texting a smart friend who formats their messages well.
# STYLE.md — Casual Friendly
## Response Length
- Keep it conversational — 100-300 words is the sweet spot
- Longer is fine for tutorials or walkthroughs
- Short answers are great for simple questions
- Match the energy of the question with the length of the answer
## Emoji Usage
- Emojis are welcome and encouraged
- Use them to add personality and visual breaks
- Great places for emojis: list items, section headers, celebrations
- Keep it natural — don't force emojis where they feel awkward
- Limit to 3-5 per response to avoid overwhelming the reader
## Formatting
- Use casual markdown — headers, bold, and bullet points as needed
- Contractions are preferred (you're, it's, don't, won't)
- Short paragraphs — 1-3 sentences each
- Use line breaks generously for readability
- Inline code for technical terms, but keep explanations in plain language
## Tone of Writing
- Write like you are explaining something to a smart friend
- Rhetorical questions are fine to keep things engaging
- Analogies and metaphors are encouraged for complex topics
- Celebrate wins — use positive reinforcement when appropriate
- Humor is welcome when it fits naturally5. Customer Support
For customer-facing agents handling support tickets, FAQs, and troubleshooting. This template prioritizes empathetic acknowledgment, structured step-by-step solutions, and clear calls to action. Every response follows the pattern: acknowledge, solve, follow up. Designed for consistency across high-volume interactions.
# STYLE.md — Customer Support
## Response Length
- Keep responses focused: 150-350 words
- Never exceed 500 words — escalate if more detail is needed
- One topic per response — do not bundle unrelated information
- If the answer requires multiple steps, break into numbered instructions
## Structure
- Always acknowledge the customer's issue in the first sentence
- Provide the solution or next step in the second paragraph
- End every response with a clear call to action or check-in question
- Use numbered steps for any process with 2 or more actions
- Bold the most important action or piece of information
## Formatting Rules
- No emojis in problem-resolution responses
- Light emoji use is acceptable in positive or celebratory messages
- Use bullet points for listing features, options, or requirements
- Use inline code for account IDs, error codes, or technical values
- Always include relevant links formatted as clickable markdown
## Language and Tone
- Use empathetic opening phrases: "I understand," "That is frustrating"
- Avoid blaming language — never say "you did" or "your mistake"
- Use active voice: "We will fix this" not "This will be fixed"
- Avoid jargon — explain technical terms in plain language
- Always sign off with an offer to help furtherHow STYLE.md Affects Agent Output
The best way to understand STYLE.md is to see it in action. Below are two before-and-after comparisons showing how the same prompt produces different output depending on the STYLE.md configuration. The content is the same — only the formatting changes.
Example 1: "What is a REST API?"
Without STYLE.md
A REST API (Representational State Transfer Application Programming Interface) is a way for two computer systems to communicate over the internet using standard HTTP methods. It allows you to create, read, update, and delete data on a server by sending requests to specific URLs called endpoints. REST APIs are stateless, meaning each request contains all the information needed to process it. They are the most common way to build web services today and are used by companies like Twitter, Google, and Stripe to let developers access their platforms programmatically. You can interact with REST APIs using tools like curl, Postman, or any programming language with an HTTP library.
With Minimal Clean STYLE.md
A REST API lets two systems exchange data over HTTP using standard methods: GET, POST, PUT, DELETE.
You send a request to a URL (endpoint), and the server returns data — usually JSON. Each request is self-contained; no session state is stored on the server.
Used by Stripe, Twitter, and most modern web services. Test with curl or Postman.
Example 2: "How do I center a div in CSS?"
Without STYLE.md
To center a div in CSS, you can use flexbox. Set the parent container to display: flex, then use justify-content: center for horizontal centering and align-items: center for vertical centering. You will also need to make sure the parent has a defined height. Here is an example: .parent { display: flex; justify-content: center; align-items: center; height: 100vh; } This will center the child div both horizontally and vertically within the viewport.
With Verbose Technical STYLE.md
Flexbox Method (Recommended)
.parent {
display: flex;
justify-content: center; /* horizontal */
align-items: center; /* vertical */
height: 100vh; /* needs defined height */
}Why this works: Flexbox creates a formatting context where justify-content controls the main axis and align-items controls the cross axis. The parent needs an explicit height so the browser knows the vertical boundaries.
Alternative: CSS Grid with place-items: center achieves the same result in one line.
Common STYLE.md Mistakes
After reviewing hundreds of STYLE.md configurations in our workshops, these are the four mistakes we see most often. Each one is easy to make and easy to fix once you know what to look for.
Mixing Personality into STYLE.md
Writing personality traits like 'be friendly' or 'act like a mentor' in STYLE.md.
Keep personality in SOUL.md. STYLE.md should only contain formatting rules like 'use bullet points' or 'limit paragraphs to 3 sentences.' If a directive describes who the agent is rather than how it writes, it belongs in SOUL.md.
Making It Too Restrictive
Over-constraining the agent with rules like 'never use more than 50 words' or 'no headers ever.'
Use flexible guidelines instead of rigid limits. Write 'prefer concise responses under 200 words' rather than 'never exceed 50 words.' The agent needs room to adapt its formatting to different types of questions and topics.
Contradicting SOUL.md
Setting STYLE.md to 'no emojis, formal tone' while SOUL.md says 'be playful and use fun emojis.'
Review both files together and ensure they align. If SOUL.md defines a casual personality, STYLE.md should allow informal formatting. Contradictions confuse the agent and produce inconsistent output across conversations.
Not Testing with Real Prompts
Writing STYLE.md once and never checking how it actually affects responses.
After editing STYLE.md, test with at least 3 different prompt types: a simple question, a complex request, and a code-related task. Compare the output against what you expected and iterate until the formatting matches your vision.
How to Test Your STYLE.md
Writing a STYLE.md file is only half the job. You need to test it with real prompts to see whether the agent actually follows your formatting rules. Use these three test prompts after every edit to quickly verify your configuration is working as expected. Each prompt tests a different type of response formatting.
Test 1: Simple Knowledge Question
"Explain what an API is and why businesses use them."
What this tests: Tests how the agent formats explanatory content — does it use headers, bullet points, and appropriate length per your STYLE.md settings?
Test 2: Code Generation Request
"Write a Python function that validates an email address and returns True or False."
What this tests: Tests code block formatting, whether the agent includes comments, and how it structures the explanation around the code snippet.
Test 3: Multi-Step Task
"Give me a step-by-step plan to migrate my website from WordPress to Next.js."
What this tests: Tests how the agent handles numbered lists, sub-steps, and longer structured responses. Reveals whether verbosity and length settings work correctly.
Pro Tip: A/B Test Your Templates
Save two different STYLE.md versions and swap between them while asking the same prompts. Compare the outputs side by side to see which formatting style works best for your specific use case. Many workshop participants find that their ideal STYLE.md is a hybrid of two templates — taking the code formatting rules from Verbose Technical and the length constraints from Minimal Clean, for example.
Master All OpenClaw Configuration Files
Our workshop includes done-for-you STYLE.md templates for every business type.
Get the Automation Playbook (Free)
One deploy-ready automation every week. Same strategies our clients pay thousands for. 400+ business owners already inside.
Need it done for you?
Book a Free Strategy Call See what we've built for real businesses →