Auto-Generate Structured Prompts with Make.com

Overview

This is a production automation solution for prompt engineering.

Input simple task descriptions to auto-generate professional-grade structured prompts:

1. Input Requirements - Input 10-word task description in Notion
2. Keyword Generation - OpenRouter generates search keywords
3. Material Scraping - Exa scrapes related web content
4. Smart Processing - GPT-4o processes to generate structured prompts
5. Save Knowledge Base - Auto-write back to Notion

Complete workflow for automatic prompt generation

---

Core Decision Factors

When choosing prompt automation solutions, consider:

• Automation Level - End-to-end automation reducing manual intervention
• Generation Quality - Structured level and practical effectiveness of prompts
• Ease of Use - Complexity of setup and daily operations
• Extensibility - Integration capability with different AI models and data sources
• Cost-effectiveness - Balance between API costs and efficiency gains

---

Technical Specifications

Specification	Value	Notes
Input Requirement	10 words	Chinese or English
Core Platform	Make.com	Workflow orchestration
Knowledge Base	Notion	Storage and triggering
Keyword Generation	OpenRouter	GPT-4o Mini
Web Scraping	Exa	Scrape prompt materials
Smart Processing	GPT-4o Mini	Generate structured prompts
Batch Capability	Hundreds to thousands	Single batch generation
Framework Count	18 types	Available prompt frameworks
Cost Reference	~$100/half year	GPT-4o Mini usage experience

---

Prerequisites

Before starting, ensure you have:

• Make.com account (free registration)
• Notion account and database
• OpenRouter API key
• Exa API key (for web scraping)

Notion Database Structure

Create prompt management database with these fields:

• Task Description (Text) - Simple task requirements
• Status (Select) - Pending/Processing/Completed
• Generated Prompt (Text) - AI-generated structured prompt
• Reference Materials (Text) - Scraped web materials (optional)

---

What are Structured Prompts

Comparison: Unstructured vs structured prompts

Structured prompts contain complete element frameworks:

Key components of prompt framework

Core Elements:

• Role Definition - AI’s professional role
• Goal Description - Specific tasks to complete
• Capability Statement - AI’s required capabilities
• Constraints - Output limits and requirements
• Workflow - Execution steps and methods
• Output Format - Expected output structure
• Reference Examples - Optional sample content

---

Workflow Architecture

Stage 1: Keyword Generation

After user inputs task description, call OpenRouter to generate search keywords:

Prompt Design:

Based on the following task requirement, generate 3-5 keywords for retrieving related prompts:

Task: {{task_description}}

Requirements:
1. Keywords should find similar task prompt templates
2. Include English keywords to expand search scope
3. Return in JSON array format

---

Stage 2: Material Scraping

Use Exa to scrape related prompt materials from web pages:

Configuration Points:

• Use generated keywords for search
• Scrape open-source prompt websites like prompts.chat
• Get multiple reference materials for processing

---

Stage 3: Smart Processing

Use GPT-4o Mini to process materials and generate structured prompts:

Prompt Design:

You are a prompt engineering expert. Based on the following task requirements and reference materials, generate a high-quality structured prompt.

Task Requirements: {{task_description}}
Reference Materials: {{reference_materials}}

Output Requirements:
1. Include role, goal, capabilities, constraints, workflow, output format
2. Language fluent and professional
3. Directly usable for AI conversations

Note: Should describe “what task to complete” rather than “generate what prompt”, otherwise AI may misunderstand.

---

Stage 4: Save Knowledge Base

Structured prompt stored in Notion

Save generated prompt back to Notion:

Considerations:

• Notion text field limits ~2000 characters
• Long text needs map and join function processing
• Update status to “Completed”

---

Batch Generation Mode

Workflow supports batch generation of hundreds of prompts:

Process:

1. Scrape base prompts from websites like prompts.chat
2. Process one by one using Iterator
3. Transform each base prompt to structured version
4. Batch save to Notion

Application Scenarios:

• Build personal prompt knowledge base
• Attract users as quality content
• Sell prompt templates on e-commerce platforms

---

Gotchas

Common pitfalls during setup:

1. Simple Model Limitations - Small models lacking reasoning only repeat execution, cannot understand modification intentions

2. Disable Auto Fallback - Must disable Automatic Fallback to avoid switching to free low-quality models

3. Notion Long Text - 2000 character limit; need map+join function to handle long text with blank lines

4. JSON Structure Break - Special characters may break JSON; use native JSON create module for escaping

5. Task Description Precision - Describe “what task to complete” not “generate what prompt”

6. Batch Consumes Operations - Batch generation consumes large amounts of Make operations; watch billing limits

---

Generation Effect Examples

Auto-generated SEO meta description expert structured prompt

Example: Viral Social Media Post Expert

• Role: Senior social media content creator
• Goal: Generate high-engagement, high-conversion viral posts
• Capabilities: Clickbait title techniques, emotional expression, trending integration
• Workflow: Analyze topic → Write title → Create body → Optimize tags
• Output Format: Title + body + tags complete post

---

Use Cases

Recommended Users

• AI Content Creators - Need high-quality structured prompts
• Prompt Engineers - Want to automate prompt production workflows
• Social Media Operators - Batch generate platform content prompts
• Efficiency Pursuers - Reduce repetitive work to focus on creativity

May Not Suit

• Users lacking basic understanding of AI tools and Make platform
• Users only occasionally needing simple prompts
• Extremely budget-limited users unwilling to use paid models

---

FAQ

Why need structured prompts?

Structured prompts contain complete elements like role, goal, constraints, workflow, etc., enabling AI to produce more professional and stable results than simple instructions.

What are the input requirements?

Only need ~10-word task descriptions (Chinese or English), like “viral social media post” or “SEO meta description expert”.

Can I batch generate?

Yes, workflow supports batch scraping materials from web pages to generate hundreds or even thousands of structured prompts at once.

Are AI model costs high?

GPT-4o Mini costs extremely low, ~$100 lasts half a year. But must disable automatic fallback to avoid switching to free low-quality models.

---

Next Steps

After mastering basics, you can try:

• Adding more prompt framework templates
• Integrating more data sources (GitHub, forums, etc.)
• Adding prompt scoring and filtering functions
• Building prompt commercialization monetization workflows

Questions? Feel free to leave comments!