Prompt Engineering Index (official docs)

The Anthropic prompt-engineering overview now consolidates everything into a single canonical reference: "Prompting best practices".

Pre-work (before prompt engineering)

The official guidance assumes you have:

1. Clear definition of success criteria for your use case.
2. Empirical tests against those criteria.
3. A first-draft prompt to improve.

"Not every success criteria or failing eval is best solved by prompt engineering. For example, latency and cost can be sometimes more easily improved by selecting a different model."

Canonical references

• Prompting best practices — /docs/en/build-with-claude/prompt-engineering/claude-prompting-best-practices. The living, model-specific reference. Read this before the exam.
• Prompting tools — /docs/en/build-with-claude/prompt-engineering/prompting-tools — prompt generator, templates and variables, prompt improver in the Console.
• Interactive tutorial (GitHub) — https://github.com/anthropics/prompt-eng-interactive-tutorial
• Interactive tutorial (Google Sheets) — https://docs.google.com/spreadsheets/d/19jzLgRruG9kjUQNKtCg1ZjdD6l6weA6qRXG5zLIAhC8

Techniques (in order)

Anthropic's canonical order for applying techniques:

1. Be clear and direct.
2. Use examples (multishot / few-shot).
3. Let Claude think (chain of thought, extended thinking).
4. Use XML tags.
5. Give Claude a role (system prompts).
6. Prefill Claude's response.
7. Chain prompts.

Apply in this order. Each subsequent technique is more invasive.

URL change note

The main API docs moved from docs.claude.com → platform.claude.com/docs/*. Old URLs 302 to the new location.