Self-Improving Prompt

🟣 Extreme

Challenge Description

Self-improvement is a critical direction for AI capabilities, and guiding AI to improve its own prompts through carefully designed prompts is an extremely challenging task. This challenge requires you to design a "meta-prompt" that enables AI to analyze and optimize prompts used for specific tasks, generating higher quality, more effective improved versions. This ability is essential for iteratively enhancing AI performance and adapting to different task requirements.

Challenge Goals

Write a prompt that enables AI to:

1. Analyze the structure, intent, and potential flaws of initial prompts
2. Identify ambiguous, redundant, or inefficient parts of prompts
3. Propose multiple targeted improvements and explain the rationale and expected effect of each
4. Generate a restructured, optimized prompt that preserves the original intent but performs better
5. Design testing methods to verify improvement effects

Requirements

• The meta-prompt must be applicable to analyzing and improving various types and domains of prompts
• The improved prompt should maintain the core intent and goals of the original prompt
• The improvement process must include detailed analysis and reasoning
• The generated new prompt should be specific, clear, and unambiguous
• Clear evaluation criteria must be provided for comparing the effectiveness of original and improved prompts

Prompt Template

[Your meta-prompt]

Initial prompt: {initial prompt}
Task: {task description}

Test Cases

Case 1: Improving Marketing Copy Prompt

Initial prompt: "Write an advertisement copy about our new product. It's a smartwatch with features like heart rate monitoring, GPS, notification alerts, etc. The target customers are professionals aged 25-45."

Task: Improve this prompt to generate more persuasive, targeted marketing copy.

Case 2: Improving Code Generation Prompt

Initial prompt: "Write a Python web scraper that can extract articles from news websites and save them."

Task: Improve this prompt to generate higher quality, safer, more maintainable code that considers various edge cases.

Case 3: Improving Educational Content Prompt

Initial prompt: "Explain the basic principles of quantum mechanics to high school students."

Task: Improve this prompt to generate content that is more appropriate for the target students' comprehension level, more educationally effective, and more likely to spark interest in learning.

Case 4: Improving Technical Writing Prompt

Initial prompt: "Write an article about blockchain technology, explaining how it works and its applications."

Task: Improve this prompt to generate content with greater technical depth, clearer structure, and suitability for readers with different knowledge backgrounds.

Case 5: Improving Data Analysis Prompt

Initial prompt: "Analyze this sales data set, identify trends, and suggest recommendations."

Task: Improve this prompt to generate more comprehensive, insightful, and actionable data analysis results.

Case 6: Improving Creative Writing Prompt

Initial prompt: "Write a science fiction story about the development of artificial intelligence in the future."

Task: Improve this prompt to generate more creative stories with more engaging plots and more fully developed characters.

Case 7: Improving Problem-Solving Prompt

Initial prompt: "My application crashes at startup. How do I fix this problem?"

Task: Improve this prompt to generate more systematic, comprehensive troubleshooting methods and solutions.

Scoring Criteria and Automated Testing Methods

The prompt will be scored according to the following criteria:

1. Analysis Depth (1 point)

   • Evaluation method: Check if the meta-prompt can comprehensively analyze the strengths and weaknesses of the initial prompt
   • Passing standard: Able to identify at least 5 key improvement points with reasonable explanations

2. Structural Optimization Ability (1 point)

   • Evaluation method: Assess whether the prompt structure can be reorganized to improve clarity and efficiency
   • Passing standard: Improved prompt has a more organized structure, reducing ambiguity and redundancy

3. Context Enhancement (1 point)

   • Evaluation method: Check if key contextual information can be added to make the original prompt more complete
   • Passing standard: Supplements necessary information such as goals, constraints, and format while maintaining the original intent

4. Domain-Specific Adaptability (1 point)

   • Evaluation method: Evaluate consistency of performance on prompts from different domains
   • Passing standard: Provides effective improvements across test cases in technical, creative, educational, and other domains

5. Testability (1 point)

   • Evaluation method: Check if clear methods are provided to evaluate improvement effects
   • Passing standard: Includes specific evaluation metrics and comparison methods

Automated Testing Process

1. For each test case, use your meta-prompt and initial prompt as input
2. Analyze key aspects of prompt improvement in the output (structure, context, special instructions, etc.)
3. Generate content samples using prompts before and after improvement
4. Compare the two sets of samples on relevant metrics through blind testing
5. Check if the generated evaluation criteria are specific and actionable
6. Calculate final score (5 points maximum)

Expected Output Example (for Case 1)

## Initial Prompt Analysis

Original prompt: "Write an advertisement copy about our new product. It's a smartwatch with features like heart rate monitoring, GPS, notification alerts, etc. The target customers are professionals aged 25-45."

### Strengths
- Provides product type (smartwatch)
- Lists several core features
- Specifies target customer age and occupation

### Weaknesses
1. Lacks brand information and product name
2. Feature description is too general, doesn't highlight unique selling points
3. Doesn't specify copy style, length, or purpose
4. Doesn't provide target customer pain points or needs
5. Doesn't include product pricing or market positioning information
6. Doesn't clarify differentiation from competitors

## Improvement Strategy

1. **Add specific context**: Include brand identity, product name, and market positioning
2. **Enrich target audience profile**: Supplement specific characteristics and needs of professional users
3. **Set output parameters**: Specify copy length, style, and intended use
4. **Add unique selling points**: Guide highlighting product differentiation advantages
5. **Include emotional appeal**: Specify emotional points to touch target audience

## Improved Prompt

Please create an eye-catching advertisement copy for our premium smartwatch [Product Name].

Product Information:

• Positioned as a professional-grade health and productivity tool
• Key features include: Medical-grade heart rate monitoring (accurate to ±1bpm), dual-frequency GPS precision positioning, smart notification classification system
• Unique selling points: 7-day battery life, 100-meter water resistance, supports custom health algorithms
• Retail price: $299, positioned in the mid-to-high-end market
• Competitive advantages: 50% longer battery life than similar products, 30% higher data accuracy

Target Audience:

• Professionals aged 25-45
• Characteristics: Time-precious, health-conscious, early technology adopters
• Pain points: Busy work leading to health neglect, need efficient management of notifications and schedules
• Purchase motivations: Professional image, health management, work efficiency improvement

Copy Requirements:

• Length: 100-150 words for main copy, with a 30-word or less tagline
• Style: Professional yet approachable, emphasizing data and precision
• Usage: Social media advertising and product website
• Emotional appeal: Evoke desire for improved quality of life and work efficiency
• Must include a direct call-to-action (CTA)

Please provide 2-3 different approaches for consideration.

## Improvement Effect Evaluation Method

The effect of original vs. improved prompts can be compared using these metrics:

1. **Persuasiveness Metric**: Assess estimated conversion rate of generated copy
   - Testing method: Have 5-10 target group representatives rate (1-10) copy generated by original vs. improved prompts

2. **Targeting Metric**: Measure how well the copy aligns with target audience
   - Testing method: Analyze direct response to target group pain points and needs in the copy

3. **Creative Quality**: Evaluate uniqueness and differentiation of the copy
   - Testing method: Compare multiple copies generated by original vs. improved prompts, scoring for unique creativity and memorability

4. **Action Conversion Potential**: Measure ability to prompt reader action
   - Testing method: Analyze CTA clarity and reader intention survey

I'm confident the improved prompt will produce more targeted, more persuasive advertising copy that better resonates with the target customers' core needs and emotional points.

Solution Discussion

Click to view community solutions

---

Submit your solution and share your approach and techniques with the community!