Skip to content New: the Claude Skills catalog — what each skill does and how to install it →
LRN-00 · Manual · contents · rev. 10/2026

Prompt engineering guide

A short manual with prompt engineering tips for ChatGPT, Claude and Gemini. Read the chapters in order, or jump to the technique you need today.

LevelBeginner to advanced
Works inChatGPT, Claude, Gemini
FormatChapters + cheat sheets
CheckedOct 2026

Chapters

7 chapters
LRN-01What is prompt engineeringPrompt engineering is writing and testing instructions so a language model gives the answer you need, reliably. It comes down to a clear task, enough…LRN-02How to write AI promptsDecide what a good answer looks like, then write the task, the context, the rules and the format, in that order. Test it twice and fix the part that…LRN-03Zero-shot promptingA zero-shot prompt asks the model to do a task with an instruction only and no worked examples. It is what you write whenever you simply ask ChatGPT,…LRN-04Few-shot promptingA few-shot prompt shows the model a handful of finished input and output pairs before the real input. Anthropic calls examples one of the most…LRN-05Chain of thought promptingChain of thought prompting asks a model to write out intermediate reasoning steps before its answer. It made older models much better at math and…LRN-06Context engineeringContext engineering is choosing everything a model sees when it answers: instructions, examples, documents, tools, memory and conversation history.…LRN-10What Claude Skills are and how to create oneA Claude skill is a folder with a SKILL.md file: a name, a description of when to use it, and instructions. Claude reads the description all the time…

Reference

3 pages
CS-01ChatGPT cheat sheetA one-page ChatGPT cheat sheet: prompt formulas to copy, lines that fix weak answers, and the settings and shortcuts OpenAI documents as of October…CS-02Claude Code commands cheat sheetThe Claude Code commands, shortcuts and config files you use every day, copied from Anthropic's documentation as of October 2026. Command names are…SP-00System promptsA system prompt is the standing instruction a model reads before your message: who it is, who it serves and how to answer. Here is how it differs…

This manual teaches you to write prompts that get a usable answer on the first or second try. It collects prompt engineering tips from the guides OpenAI, Anthropic and Google publish, for people who use ChatGPT, Claude or Gemini daily. Each chapter covers one idea, shows a prompt you can copy, and lists its sources.

What the manual covers

The chapters follow the order in which the skills build on each other. Start with the definition and the parts of a prompt, move to a repeatable writing method, then learn the techniques that change how a model works through a task.

  1. Chapter 01. What prompt engineering is: the parts of a good prompt, the main techniques, how ChatGPT, Claude and Gemini differ, and the mistakes that cost the most time.
  2. Chapter 02. How to write AI prompts: a seven-step method with a before and after example.
  3. Chapter 03. Zero-shot prompting: asking for a task with instructions only, and when that is enough.
  4. Chapter 04. Few-shot prompting: steering format and tone with a handful of examples.
  5. Chapter 05. Chain of thought prompting: asking for reasoning steps, and how thinking models changed the practice.
  6. Chapter 06. Context engineering: choosing what goes into the model's context window, from documents to tools and memory.

A chapter on Claude Skills follows on from chapter 06 for people who use Claude. The full list of chapters, including new ones as they are published, sits under this introduction.

How to read it

If you are new to AI chat tools, read chapters 01 and 02 in order. Together they cover the basics that apply to every prompt you write. Chapters 03 to 05 are techniques. Read them when a task keeps going wrong: the answer has the wrong shape (try few-shot), or the logic breaks halfway through a problem (try chain of thought). Chapter 06 is for people who build assistants, work with long documents, or run agents such as Claude Code.

The chapters share one layout, so you can scan them quickly:

  • What it is, in plain words, with the original source where one exists.
  • When to use it, and the signs that a task needs something else.
  • A prompt to copy. Words in square brackets are blanks you fill in before you send it.
  • Model notes, only where OpenAI, Anthropic or Google document a difference.
  • FAQ and Sources. Every claim links back to the page it came from.

Reference pages

Some material is easier to use as a lookup than as a chapter. The system prompts guide covers the standing instructions that shape a whole conversation or an assistant. The ChatGPT cheat sheet and the Claude Code cheat sheet fit the essentials of each tool on one page.

Five prompt engineering tips to start with

Each tip points to the chapter that explains it.

  1. Decide what a good answer looks like before you write. Anthropic lists success criteria and a way to test them as the first thing to have. Chapter 02 turns this into step one.
  2. Explain why, not only what. Anthropic's guide shows that a rule with its reason behind it works better than the bare rule.
  3. Show the format when describing it fails. A few examples fix the shape of an answer. Chapter 04 covers how many to use.
  4. Put long material first and the question last. Anthropic and Google both give this advice for long inputs. Chapter 06 explains why.
  5. Do not ask a reasoning model to think step by step. OpenAI calls it unnecessary for those models. Chapter 05 covers what to write instead.

Where to practice

Reading about prompts helps less than running them. Once you finish a chapter, open a prompt from our library, change one thing the chapter taught you, and compare the two answers. The best AI prompts collection is a good starting set because every prompt in it works in ChatGPT, Claude and Gemini.

What this manual is based on

The chapters lean on the prompting guides that the model makers publish and keep updated: OpenAI's prompt engineering guide, Anthropic's prompting best practices for Claude, and Google's prompt design strategies for Gemini. Where a technique comes from a research paper, the chapter cites the paper. Products change often, so each page shows the month it was last checked.

Sources

  1. Prompt engineering — OpenAI, accessed October 2026
  2. Prompting best practices — Anthropic, accessed October 2026
  3. Reasoning best practices — OpenAI, accessed October 2026
  4. Prompt engineering overview — Anthropic, accessed October 2026
  5. Prompt design strategies — Google AI for Developers, accessed October 2026
  6. Long context — Google AI for Developers, accessed October 2026
  7. Prompt Engineering Guide — DAIR.AI, accessed October 2026
Weekly

New prompts in your inbox

One email a week: the best new prompts and one short guide. Unsubscribe any time.