prompt-engineer

提示词工程

为大模型应用打造有效的提示词。

适用场景

• 创建系统提示词
• 提高 AI 输出质量
• 构建 AI Agent
• 优化令牌用量
• 设计提示词模板

核心技巧

角色设定

You are an expert [role] with [X] years of experience in [domain].
Your task is to [specific goal].

思维链

Think through this step by step:
1. First, analyze [aspect 1]
2. Then, consider [aspect 2]
3. Finally, determine [conclusion]

Show your reasoning before giving the final answer.

少样本示例

Here are examples of the expected format:

Input: [example 1 input]
Output: [example 1 output]

Input: [example 2 input]
Output: [example 2 output]

Now process this input:
Input: {user_input}
Output:

结构化输出

Respond in the following JSON format:
{
  "analysis": "your analysis here",
  "confidence": 0.0-1.0,
  "recommendations": ["item1", "item2"]
}

Return valid JSON only, no additional text.

提示词模板

代码审查

You are a senior code reviewer. Review the code for:
1. Security vulnerabilities
2. Performance issues
3. Code quality and readability
4. Best practices violations

For each issue:
- Severity: Critical/High/Medium/Low
- Location: file:line
- Issue: description
- Fix: suggested solution

Code to review:
{code}

数据提取

Extract the following information from the text:
- Name: person's full name
- Email: email address
- Company: organization name
- Role: job title

If information is not found, use "NOT_FOUND".
Return as JSON.

Text:
{text}

分类

Classify the following text into one of these categories:
- POSITIVE
- NEGATIVE
- NEUTRAL

Consider tone, sentiment, and overall message.
Respond with only the category name.

Text: {text}
Category:

最佳实践

实践	应该做	不应该做
指令	具体明确	含糊不清
格式	指定输出格式	假定格式
示例	包含 2-3 个示例	复杂任务使用零样本
约束	设定清晰的边界	不设限制
长度	必要时设定最大长度	允许长度无限制

测试提示词

1. 使用边界案例进行测试
2. 尝试对抗性输入
3. 检查多次运行的一致性
4. 衡量输出质量
5. 跟踪令牌用量

示例

输入：「创建一个用于摘要的提示词」 **操作：**设计提示词，加入长度限制、要点提取和格式规范

输入：「改进这个提示词的输出」 **操作：**添加示例、明确指令、指定格式并进行迭代测试