How to Engineer High-Level Prompts: Examples and Improvements

Enhanced Prompt:

Title: How to Engineer High-Level Prompts: Examples and Improvements

Objective:
To create a poorly-written AI prompt and then systematically reengineer it into a high-level prompt that meets industry best practices and standards. The process should also include defining quality expectations, incorporating technical considerations, addressing potential challenges, and ensuring scalability and performance optimization.

Deliverables:

1. A clear example of a poorly-written AI prompt and a corresponding well-engineered high-level prompt.
2. An analysis of the issues present in the poorly-written prompt, including specific ways to improve clarity, specificity, and technical detail.
3. A reengineered high-level prompt that incorporates best practices, including quality assurance metrics and performance benchmarks.
4. A detailed explanation of the improvements made, such as:
   • Adding specificity to the objective.
   • Including actionable instructions.
   • Incorporating meta instructions for efficiency.
   • Specifying acceptable ambiguity and precision requirements.
   • Defining success criteria, such as pass rates, accuracy, or BLEU scores.
5. Guidance on the challenges of prompt engineering and the solutions employed.
6. A discussion of scalability, including suggestions to break down complex tasks into manageable components.
7. Technical specifications, such as:
   • Recommended AI frameworks (e.g., Python, LangChain, PyTorch).
   • Tools or technologies for automating or improving prompt engineering (e.g., prompt templates, AI simulators).
   • Evaluation metrics and tools.
8. Considerations for security and performance, such as avoiding overly broad prompts, optimizing for computational efficiency, and ensuring numerical/quantitative answers are provided where applicable.

Success Criteria:

• The poorly-written prompt should remind the user of common pitfalls in prompt engineering.
• The high-level prompt should align with specific metrics or K