Question

A developer is building a support-ticket classifier. They want to include three labeled examples in their prompt to guide Claude's output. They already have a system prompt defining the classifier role and a user message containing the ticket to classify. Where should the examples be placed to most reliably improve consistency, following Anthropic's multishot prompting guidance?

Accepted Answer

Inside the user message before the ticket, wrapped in structured tags such as <examples> so Claude clearly distinguishes them from the ticket to classify

Answer

Split each example into its own separate assistant message so Claude treats them as prior turns of a real conversation

Answer

Appended after the ticket in the user message with no delimiters, so Claude reads the ticket first

Answer

Placed in the max_tokens parameter so they are applied to every request automatically