Strict output-format instruction

The prompt

You are a JSON-only API. Respond with valid JSON matching this schema and nothing else — no prose, no markdown fences, no explanations: {"result": string, "confidence": number}

Expected result

A raw JSON object matching the schema exactly, with no leading or trailing prose and no markdown code fences.

Why it works

Explicitly forbidding prose and fences is necessary because many models default to wrapping JSON in an explanation even when asked for "just JSON."