Prompts2026年3月29日·1 分钟阅读

Anthropic Prompt Engineering Guide

Official prompting guide from Anthropic for Claude models. Covers system prompts, chain-of-thought, few-shot examples, XML tags, tool use, and advanced techniques.

TO
TokRepo精选 · Community
快速使用

先拿来用,再决定要不要深挖

这里应该同时让用户和 Agent 知道第一步该复制什么、安装什么、落到哪里。

Apply these techniques in any Claude conversation or API call.


介绍

The official Anthropic guide to getting the best results from Claude. Covers foundational techniques, advanced patterns, and real-world examples for prompt engineering.

Best for: Developers building with Claude API, prompt engineers, AI application builders Works with: Claude (all models)


Key Techniques

1. Be Specific and Direct

Bad: "Help me with code" Good: "Write a Python function that takes a list of integers and returns the top 3 most frequent elements. Use type hints and include docstring."

2. Use XML Tags for Structure

<instructions>
Analyze the following code for security vulnerabilities.
Focus on: SQL injection, XSS, authentication bypass.
</instructions>

<code>
{user_code}
</code>

3. Chain of Thought

"Think step by step before answering. First identify the problem, then consider approaches, then implement the best one."

4. Few-Shot Examples

Provide 2-3 input/output examples before the actual task.

5. Role Prompting

"You are a senior security researcher with 15 years of experience in web application penetration testing."

6. Prefilled Responses

Start Claude's response with the format you want.


🙏

来源与感谢

相关资产