Claude
Claude prompt engineering: the techniques Anthropic documents and how a team applies them
A plain guide for teams who want more reliable results from Claude. Every technique comes from Anthropic's own documentation as read on 10 November 2026. Model advice changes often, so check the current page for your model before you rely on it.
Claude prompt engineering is the practice of writing instructions that make Claude give the result you want. Anthropic's documentation covers techniques that include being clear and direct, adding context that explains why, using a few well-made examples, structuring prompts with XML tags, and giving Claude a role in the system prompt. It also says to define success criteria and test against them first, because not every failing result is best fixed by changing the prompt. Our view: treat a prompt like a work instruction that a new colleague could follow. Sources read 10 November 2026.
plain answer
Claude prompt engineering: what is it?
Prompt engineering is writing and refining the instructions you give a language model so that it produces the result you need. Anthropic's prompt engineering overview points to its prompting best practices page as the living reference for the techniques, which it lists as clarity, examples, XML structuring, role prompting, thinking and prompt chaining.
Anthropic's overview assumes you already have a clear definition of success criteria for your use case, some ways to test against those criteria, and a first draft prompt to improve. It says to establish those first.
Sources: Anthropic, Prompt engineering overview (https://platform.claude.com/docs/en/build-with-claude/prompt-engineering/overview) and Anthropic, Prompting best practices (https://platform.claude.com/docs/en/build-with-claude/prompt-engineering/claude-prompting-best-practices), both read 10 November 2026.
how to
What are the main Claude prompting techniques?
Be clear and direct. Anthropic says Claude responds well to clear, explicit instructions, and suggests thinking of Claude as a brilliant but new employee who lacks context on your norms and workflows. Its test: show your prompt to a colleague with minimal context; if they would be confused, Claude will be too.
Add context. Anthropic says explaining why an instruction matters helps Claude deliver more targeted responses.
Use examples. Anthropic calls examples one of the most reliable ways to steer output format, tone and structure, and says to include 3 to 5 that are relevant, diverse and wrapped in example tags.
Structure with XML tags. Anthropic says tags help Claude parse prompts that mix instructions, context, examples and variable inputs.
Give Claude a role. Anthropic says setting a role in the system prompt focuses Claude's behaviour and tone, and even one sentence makes a difference.
Source: Anthropic, Prompting best practices (https://platform.claude.com/docs/en/build-with-claude/prompt-engineering/claude-prompting-best-practices), read 10 November 2026.
limits of prompting
When is prompt engineering not the answer?
Anthropic says not every success criterion or failing test is best solved by prompt engineering. Its example is that latency and cost can sometimes be improved more easily by choosing a different model.
Anthropic's best practices page also says that where a technique names a specific model, you should treat it as measured on that model and re-check it against your own tests before applying it to another. Our reading: keep a small set of real test cases for each prompt your team relies on, and re-run them when the model changes.
Sources: Anthropic, Prompt engineering overview (https://platform.claude.com/docs/en/build-with-claude/prompt-engineering/overview) and Anthropic, Prompting best practices (https://platform.claude.com/docs/en/build-with-claude/prompt-engineering/claude-prompting-best-practices), both read 10 November 2026.
kriv role
Where Kriv helps
Kriv AI trains teams to use Claude and helps organisations set it up so staff work with it inside agreed rules. For prompting, that means working from your team's real tasks: writing shared prompts, building a small set of test cases, and agreeing which outputs a person must check before anyone uses them.
Kriv did not write Anthropic's documentation. Scope and price are agreed after a short discovery call.
limits
What this page could not confirm
This page rests on two Anthropic documentation pages, as read on 10 November 2026, neither of which shows a publication date. We did not test the techniques on a specific task, did not read the model-specific pages, and did not compare Claude with other models. Anthropic updates this guidance with each model release, so check the current pages for the model you use.
Straight answers
Frequently asked questions about Claude prompt engineering: the techniques Anthropic documents and how a team applies them
What is Claude prompt engineering?
It is writing and refining instructions so Claude gives the result you need. Anthropic's documentation lists clarity, examples, XML structuring, role prompting, thinking and prompt chaining among the techniques (read 10 November 2026).
How many examples should I put in a Claude prompt?
Anthropic's documentation says to include 3 to 5 examples for best results, and to make them relevant, diverse and wrapped in example tags (read 10 November 2026).
Why use XML tags in a Claude prompt?
Anthropic says XML tags help Claude parse prompts that mix instructions, context, examples and variable inputs, which reduces misinterpretation (read 10 November 2026).
Should I give Claude a role?
Anthropic says setting a role in the system prompt focuses Claude's behaviour and tone, and that even a single sentence makes a difference (read 10 November 2026).
Can prompt engineering fix every poor result?
No. Anthropic says not every failing test is best solved by prompt engineering, and that a different model can sometimes improve latency and cost more easily (read 10 November 2026).
Does Kriv train teams on prompting Claude?
Kriv AI trains teams to use Claude. See our prompt writing course and Claude training pages for what is offered; price is agreed after a short discovery call.
Talk to the team that would do the work
Bring your requirements to a working session with the person who'll actually deliver.
Book a Discovery Call