Claude Prompt Fieldbook

Fast paths

Choose what you’re doing

Use it now

Pocket library

Prompt patterns

Task Level Scope

Favorites

Prompt builder

Answer four short questions. Your inputs stay in this browser and are inserted into a reusable prompt.

Step 1 of 5Task

01 · Task

What should Claude do?

BuildCreate or change code DebugFind and fix a failure ReviewInspect code for issues ExplainLearn a concept or codebase

05 · Ready

Your prompt is ready.

Place stable behavior in a system prompt when possible; keep the changing task and context in the user prompt.

Generated prompt

Decision guide

Fix a result

Choose the closest symptom. You’ll get a focused diagnosis, a prompt patch, and a fallback path.

Primary sources

Verified against Anthropic documentation

This fieldbook summarizes and restructures official guidance; it is not affiliated with Anthropic. Links require a network connection.

01

Prompting Claude Fable 5Model-specific behavior, long runs, effort, memory, safety

02

Prompting best practicesClarity, examples, XML, long context, tools, agent workflows

03

Models overviewAvailability, model IDs, capabilities, compatibility

04

EffortSupported levels and capability/latency/token trade-offs

05

Refusals and fallbackFable 5 refusal stop reason and fallback handling

Write prompts that make hard coding work feel tractable.

Choose what you’re doing

Use it now

Prompt patterns

No matching patterns

Favorites

Recently viewed

Prompt builder

What should Claude do?

What does “done” look like?

Give Claude the boundaries.

How should Claude work and report?

Your prompt is ready.

Fix a result

Verified against Anthropic documentation

Quick help

Prompt anatomy

Syntax

Fast fixes

Safety checkpoint

Keyboard