4 min read

The Complete Guide to Prompt Engineering Markup

Written by

IV

Creator

Published on

7/9/2025

A Self-Referencing Tutorial with Live Examples

Meta-note: This tutorial demonstrates prompt engineering techniques by using them in its own construction. Each technique is explained and immediately exemplified in the tutorial’s structure itself.

Foundation: Why Structure Matters

When crafting prompts, unclear instructions lead to inconsistent outputs. This tutorial itself follows structured markup principles—notice how I’m using:

  • Clear headers to separate sections
  • Bold emphasis for key concepts
  • Consistent formatting throughout
    Poor prompt: "Write something about cats make it good"
    Better prompt: "Write a 200-word informative article about domestic cats, focusing on their behaviour and care requirements."

Level 1: Basic Markdown Structure

Headers Create Hierarchy

# Primary Instruction (H1)
## Task Breakdown (H2)
### Specific Requirements (H3)
Live example: Notice how this tutorial uses exactly this header structure—each section builds logically on the previous one.

Emphasis and Lists

Bold text highlights critical requirements:
Required: Response must be exactly 500 words
Optional: Include relevant examples
Forbidden: Do not use technical jargon
Numbered lists create clear sequences:

  1. Analyse the provided text
  2. Identify key themes
  3. Summarise findings
  4. Provide recommendations
    Self-reference: This tutorial follows this exact pattern—each section builds systematically on the previous one.

Level 2: Content Boundaries and Examples

Code Blocks for Clarity

Triple backticks separate different content types:

The quick brown fox jumps over the lazy dog.
Analysis: This pangram contains all 26 letters of the English alphabet.
- Word count: 9 words
- Character count: 43 characters
- Unique letters: 26

Tutorial demonstration: I’m using this exact technique throughout—each code block clearly separates example content from instructional text.

Quotation Marks for Literal Text

When you need exact reproduction:
Respond with exactly: "Analysis complete. Results below."
Then provide your analysis.

Level 3: XML-Style Semantic Tags

Creating Logical Sections

You are a professional editor reviewing academic papers.

<input_document>
[Document content goes here]
</input_document>

<evaluation_criteria>

  • Clarity of argument
  • Evidence quality
  • Citation accuracy
  • Writing style
    </evaluation_criteria>

<output_requirements>
Provide a structured review with ratings (1-10) for each criterion.
</output_requirements>

Meta-demonstration: This tutorial could be restructured using XML tags:
<tutorial_section id="xml_tags">
<learning_objective>Understand semantic markup in prompts</learning_objective>
XML tags provide clear semantic meaning...
[Live example content]
</tutorial_section>

Referencing Sections

XML tags allow you to reference specific parts:
Follow the guidelines in <evaluation_criteria>.
Format your response according to <output_requirements>.
If the document doesn't meet standards in <evaluation_criteria>, explain why.

Level 4: Advanced JSON Schema

Structured Data Specification

{
"task": "content_analysis",
"input": {
"type": "blog_post",
"word_count": 1500,
"topic": "sustainability"
},
"analysis_framework": {
"readability": {
"metrics": ["flesch_reading_ease", "sentence_length", "paragraph_structure"],
"target_score": "60-70"
},
"engagement": {
"elements": ["hook", "storytelling", "call_to_action"],
"rating_scale": "1-10"
},
"seo_factors": {
"keyword_density": "1-3%",
"meta_elements": ["title", "description", "headers"]
}
},
"output_schema": {
"summary": "string (max 100 words)",
"scores": {
"readability": "number",
"engagement": "number",
"seo": "number"
},
"recommendations": {
"immediate": ["array of strings"],
"long_term": ["array of strings"]
},
"confidence_level": "number (0-100)"
}
}
Self-referential example: This tutorial itself could be analysed using this schema—it has clear structure (readability), engaging examples (engagement), and organised headers (SEO factors).

Template Variables in JSON

{
"task": "{{TASK_TYPE}}_analysis",
"target_audience": "{{USER_EXPERTISE_LEVEL}}",
"constraints": {
"word_limit": "{{MAX_WORDS}}",
"tone": "{{REQUIRED_TONE}}",
"format": "{{OUTPUT_FORMAT}}"
}
}

Level 5: Multi-Layered Hybrid Approaches

Combining All Techniques

# Document Enhancement System

## Core Instructions
Senior content strategist with 10+ years experience
<primary_task>Enhance provided content for maximum impact</primary_task>

## Input Processing
Document type: {{DOCUMENT_TYPE}}
Current word count: {{CURRENT_LENGTH}}
Target audience: {{AUDIENCE_PROFILE}}
Distribution channel: {{PLATFORM}}
## Enhancement Framework
{
"content_analysis": {
"structure": "assess logical flow and organisation",
"clarity": "evaluate language accessibility",
"engagement": "measure hook effectiveness and retention"
},
"optimisation_strategy": {
"IF platform == 'social_media'": "prioritise brevity and visual appeal",
"ELSE IF platform == 'academic'": "focus on rigour and citations",
"ELSE": "balance accessibility with authority"
}
}

## Quality Assurance

  • Mandatory: Preserve author's voice and intent
  • Enhanced: Improve clarity without changing meaning
  • Optimised: Adapt for specified platform requirements

## Output Specification
<enhanced_content>
[Revised content here]
</enhanced_content>

<change_log>

  • List of modifications made
  • Rationale for each change
  • Impact assessment
    </change_log>
    Tutorial meta-analysis: This tutorial demonstrates this hybrid approach—it combines markdown headers, XML-style tags, JSON schemas, and conditional logic all within a coherent structure.

Advanced Techniques: Conditional Logic and Context Management

Conditional Processing

CONTEXT_CHECK:
IF user_expertise == "beginner":
Use simple language and provide background explanations
ELSE IF user_expertise == "intermediate":
Assume basic knowledge but explain advanced concepts
ELSE IF user_expertise == "expert":
Use technical terminology and focus on nuanced insights
ENDIF

CONTENT_ADAPTATION:
IF content_type == "tutorial":
Include step-by-step examples
ELSE IF content_type == "reference":
Provide comprehensive but concise information
ENDIF

Context Variables and State Management

{
"conversation_context": {
"previous_topics": ["{{TOPIC_HISTORY}}"],
"user_preferences": {
"detail_level": "{{PREFERRED_DEPTH}}",
"example_types": ["{{EXAMPLE_PREFERENCES}}"],
"format_style": "{{OUTPUT_STYLE}}"
},
"session_state": {
"current_focus": "{{ACTIVE_TOPIC}}",
"completed_sections": ["{{FINISHED_TOPICS}}"]
}
}
}

Expert-Level: Self-Modifying and Adaptive Prompts

Dynamic Prompt Evolution

## Adaptive Learning Framework

<initial_assessment>
Analyse user's first response to determine:

  • Technical knowledge level
  • Preferred communication style
  • Specific interests within the topic
    </initial_assessment>

<prompt_modification_rules>
BASED ON initial_assessment:
UPDATE explanationdepth TO match userknowledge_level
ADJUST examplecomplexity TO usercomprehension_signals
MODIFY interactionstyle TO usercommunication_preferences
</prompt_modification_rules>

<continuous_calibration>
THROUGHOUT conversation:
MONITOR userengagementindicators
ADAPT responselength BASED ON userattention_signals
REFINE exampleselection USING userfeedback_patterns
</continuous_calibration>

Self-Referencing Validation

## Prompt Quality Assurance

<self_check_protocol>
BEFORE responding:

  1. Verify all are internally consistent
  2. Ensure actually demonstrate stated principles
  3. Confirm <output_format> matches
  4. Validate that conditional logic in IF/ELSE statements is complete </self_check_protocol>

<meta_validation> This tutorial demonstrates self-referencing by:

  • Using the markup techniques it teaches
  • Providing live examples within its own structure
  • Following the progressive complexity it advocates
  • Implementing the quality assurance it recommends </meta_validation>

Practical Application: Building Your Own Structured Prompts

Template Framework

# {{YOUR_TASK_NAME}}

## Context Setting {{DEFINEAIPERSONA}} {{SPECIFYSUBJECTAREA}} {{LIST_LIMITATIONS}}

## Input Specification Input type: {{DATA_TYPE}} Format: {{INPUT_FORMAT}} Quality requirements: {{QUALITY_STANDARDS}} ## Processing Instructions { "analysis_steps": [ "{{STEP_1}}", "{{STEP_2}}", "{{STEP_3}}" ], "quality_checks": { "accuracy": "{{ACCURACY_CRITERIA}}", "completeness": "{{COMPLETENESS_STANDARDS}}", "consistency": "{{CONSISTENCY_REQUIREMENTS}}" } }

## Output Requirements {{OUTPUT_STRUCTURE}} {{WORD_COUNT_RANGE}} {{COMMUNICATION_STYLE}}

## Validation Protocol

  • Self-check: Does output meet all specifications?
  • Quality assurance: Are examples relevant and accurate?
  • Completeness: Have all requirements been addressed?

Common Pitfalls and Solutions

Inconsistent Markup

Problem: Mixing different markup styles without clear boundaries ❌ Poor approach: # Main heading Do this task Bold requirement but also code style emphasis {json: "mixed with markdown"} Solution: Choose consistent markup hierarchy ✅ Better approach: # Main Instructions ## Task Breakdown

  • Use consistent formatting
  • Maintain clear hierarchy
  • Separate different content types

Over-Specification

Problem: Too many constraints that conflict with each other ❌ Problematic: { "length": "exactly 500 words", "requirements": ["comprehensive coverage", "detailed examples", "brief summary"], "constraints": ["no technical jargon", "expert-level analysis"] } Solution: Prioritise requirements and resolve conflicts ✅ Improved: { "primary_requirement": "comprehensive analysis in 500 words", "secondary_goals": ["clear examples", "accessible language"], "flexibility": "adjust detail level to maintain word count" }

Tutorial Conclusion: Meta-Learning Check

Self-Assessment Questions

This tutorial has demonstrated its own principles. Can you identify:

  1. Markdown usage: How many header levels were used consistently?
  2. XML tags: Which semantic sections were created?
  3. JSON schemas: What structured data was specified?
  4. Code blocks: How were examples separated from instructions?
  5. Self-reference: Where did the tutorial analyse its own structure?

Implementation Checklist

When creating your own structured prompts:

  • Clear hierarchy: Headers progress logically from general to specific
  • Consistent formatting: Same markup style throughout
  • Semantic boundaries: Different content types are clearly separated
  • Self-validation: Prompt includes quality checks
  • Adaptive elements: Accounts for different user needs
  • Live examples: Demonstrates rather than just describes

Next Steps

Immediate application: Take an existing prompt you use and restructure it using these techniques. Progressive enhancement: Start with basic markdown, then add XML tags, then incorporate JSON schema as needed. Continuous improvement: Test your structured prompts and refine based on output quality.

This tutorial has demonstrated advanced prompt engineering by using every technique it teaches within its own construction—a practical example of self-referencing documentation that serves as both instruction and implementation guide.

Latest

More from the site

View all posts