BeginnerИнженерияclaude

Anthropic's Prompt Engineering Interactive Tutorial

Anthropic's official interactive tutorial: from basic prompt structure to complex prompts from scratch. 10 chapters with exercises, examples, and real patterns from the Claude team.

4modules

10lessons

200 mintotal time

Everyone starting to work with Claudeaudience

Module 1

Structure and Clarity Foundations

How the Messages API works, how to write clear instructions, and why Claude needs explicit context.

Basic Prompt Structure

Messages API, required parameters (model, max_tokens, messages), alternating user/assistant roles, and system prompts.

20 min

Being Clear and Direct

The Golden Rule of Prompting, how to eliminate preambles, and how to make Claude commit to a single answer instead of hedging.

20 min

Module 2

Roles, Data, and Formatting

Role prompting, XML tags for separating data from instructions, output format control, and prefilling responses.

Assigning Roles (Role Prompting)

How a role in the system prompt changes tone and output quality, from a cat to a logic bot, and why roles help with math and reasoning.

20 min

Separating Data and Instructions

Prompt templates with variables, XML tags as separators, why 'Dear Claude' in the response signals mixed instructions and data.

20 min

Formatting Output and Speaking for Claude

XML tags in output for reliable parsing, prefill (speaking for Claude), enforcing JSON via prefill with '{', stop_sequences to save tokens.

20 min

Module 3

Thinking, Examples, and Hallucinations

Step-by-step thinking via XML tags, few-shot prompting, and techniques for reducing hallucinations.

Precognition: Thinking Step by Step

Why 'thinking out loud' is more accurate, examples with brainstorm and positive/negative-argument tags, and the ordering effect.

20 min

Using Examples (Few-Shot Prompting)

Zero-shot vs few-shot, how examples set the tone for a parent bot, extracting structured data through format demonstration.

20 min

Avoiding Hallucinations

Two main methods: giving Claude an out ('only if you know for certain') and requiring a quote from the document before answering.

20 min

Module 4

Complex Prompts and Chaining

Assembling a complex prompt from all learned elements and the prompt chaining technique for iterative improvement.

Complex Prompts from Scratch

20 min

Chaining Prompts

Iterative improvement through multi-turn dialogue: fixing non-words, improving a story, extracting names and alphabetizing them.

20 min