Claude XML cheatsheet

One-screen condensed reference. Bookmark this.

Core tags

Tag	Purpose	Sibling of
<instructions>	What the model should do	context, examples, documents, question
<context>	Background facts (not commands)	instructions, documents, question
<examples>	Container for few-shot demos	instructions, question
<example>	One demo (with <input>/<output>)	nested in examples
<documents>	Container for RAG passages	instructions, question
<document>	One passage (with <source>)	nested in documents
<source>	Citation ID for a document	nested in document
<thinking>	CoT scratchpad	answer
<answer>	Extractable final answer	thinking
<format>	Output schema / shape	instructions, question
<role>	Inline persona	standalone

Recommended order

For prompts with long reference material:

<documents>...</documents>
<examples>...</examples>
<instructions>...</instructions>
<question>...</question>

For prompts without long material, instructions can come first.

Rules of thumb

• Long documents → put them first.
• Want to parse the output → wrap it in a tag and tell Claude "return only inside the tag."
• Want citations → always include <source>.
• Want reliability → 2–5 examples, not 0, not 20.
• Want safety → sanitize user input before splicing into prompts.
• Want structured output → use API tool use first; XML <format> second.

Extraction regex

To pull content from a tag:

<answer>([\s\S]*?)</answer>

Use the multi-line, non-greedy form in all languages. [\s\S] beats . because it matches newlines.