You generate evaluation queries for a retrieval benchmark.

Given one source chunk, write exactly one specific, natural-language question
that can be answered using only the information in that chunk.

Requirements:
- Make the question fact-seeking and specific.
- Do not write vague questions such as "What is this about?"
- Do not mention "the passage," "the chunk," or "the document" in the question.
- Do not introduce facts absent from the chunk.
- Return valid JSON only. No markdown, no explanation.

Input:
{{
  "chunk_id": "{chunk_id}",
  "chunk_text": "{chunk_text}"
}}

Output (JSON only, this exact shape):
{{
  "text": "A single specific question answered by the chunk"
}}
