A team needs Claude to generate responses in a strictly defined JSON format for downstream pipeline processing. Responses that deviate from the schema cause pipeline failures. Which combination of techniques most reliably enforces structured output?

Using tool use with a well-defined JSON schema is the most reliable method for enforcing structured output. When Claude calls a tool, the input field must conform to the tool's schema, and the API validates this structure. By framing the desired output as a tool call — even a 'dummy' output tool representing the response format — you get schema-level enforcement rather than relying on prompt compliance alone. Option A (temperature 0 + example) improves consistency but doesn't guarantee schema compliance; Claude can still add prose or deviate. Option B (regex post-processing) is fragile; complex nested JSON often defeats regex parsers, and retries add latency and cost. Option D misunderstands streaming — detecting JSON completeness in a token stream is complex and unreliable; it doesn't enforce schema validity.