Chapter 8 · Basic Prompt Elements and Instruction Design
Chapter goals: Master the basic building blocks of prompts, learn to design clear and effective instructions, and understand the impact of instruction design on model output.
8.1 Five Elements of a Prompt
A complete prompt typically contains the following elements:
| Element | Purpose | Example |
|---|---|---|
| Role | Set the model's "persona" | "You are a professional Python engineer" |
| Task | Clarify what needs to be done | "Analyze the performance bottlenecks in this code" |
| Context | Provide background information | Code snippets, doc links, domain knowledge |
| Constraints | Limit the output scope | "Answer in Chinese", "No more than 200 words" |
| Format | Specify output structure | JSON, Markdown, list, etc. |
8.2 Instruction Design Principles
1. Clear and Specific
Avoid vague expressions; use concrete, executable instructions:
python
# ❌ Bad example
"Write something for me"
# ✅ Good example
"Write a 500-word technical blog post introducing best practices for Python async programming, aimed at developers with some background"2. Structured Organization
Use delimiters and structured formats:
python
prompt = """
# Role
You are a technical documentation expert.
# Task
Organize the following technical notes into standardized API documentation.
# Input
{notes}
# Requirements
1. Use Markdown format
2. Include a parameter description table
3. Provide at least one code example
4. Language: English
# Output Format
## API Name
## Description
## Parameters
## Example
"""3. Positive Phrasing
Prefer positive instructions over negative ones:
python
# ❌ Avoid negative phrasing
"Don't use complex terminology"
# ✅ Use positive phrasing
"Use simple and easy-to-understand language"8.3 Common Instruction Patterns
1. Role-Playing Pattern
python
role_prompt = """
You are a system architect with 10 years of experience.
Please analyze the pros and cons of the following design from an architectural perspective:
{design_doc}
"""2. Step-by-Step Reasoning Pattern
python
step_by_step = """
Please solve this problem step by step:
1. First understand the core of the problem
2. List known conditions and constraints
3. Derive the solution
4. Verify the correctness of the result
Problem: {problem}
"""3. Comparative Analysis Pattern
python
compare_prompt = """
Please compare the following two approaches:
- Approach A: {option_a}
- Approach B: {option_b}
Analyze from cost, performance, and maintainability perspectives, and give a recommendation.
"""8.4 Tips for Debugging Prompts
When the model output is unsatisfactory, try the following:
- Add more context: provide more detailed background information
- Adjust constraints: tighten or loosen restrictions
- Change the phrasing: rewording may produce better results
- Use examples: few-shot examples effectively guide the model
- Decompose the task: break complex tasks into sub-steps
Chapter Summary
- Prompts contain five elements: role, task, context, constraints, format
- Instruction design should be clear, structured, and use positive phrasing
- Master common patterns: role-playing, step-by-step reasoning, comparative analysis
- When debugging, try from five angles: context, constraints, phrasing, examples, decomposition
🧪 随堂测验
点击你认为正确的选项。答错时会展示正确答案与原因解析。
1. Which of the following is NOT one of the five elements of a prompt?
2. Which instruction design is the clearest?
3. When debugging a prompt, which method is most effective?
4. What is the main purpose of using delimiters?
🛠️ Hands-on Practice
- Design a complete prompt for a "code review assistant" containing all five elements, and test the effect difference across different phrasings.
- Prompt the same task using both "direct question" and "step-by-step reasoning" approaches, and compare output quality.
- Implement a prompt debugging tool: record outputs across different parameter combinations and automatically generate improvement suggestions.