QuestionQ124
Claude Models, Prompting & Context EngineeringYou are categorizing token-management tactics by the request-lifecycle stage at which each applies.
For each tactic, select its applicable request-lifecycle stage.
Select
Persist the validated output for downstream consumption and audit.
Order the prompt sections so cacheable content sits before per-request content.
Summarize prior conversation history when full history is no longer needed.
Validate the model's structured output against the expected schema.
Move stable repeated content into a cacheable prefix at the head of the prompt.
Trim retrieved passages to spans relevant to the user's question.
Community Discussion