把固定提问用于三场访谈，按来源区分意见变化
本页展示随包原生输入和验收参考。目标产品运行状态：未运行。表格与图片不代表产品已经生成的结果。
input.csv / practice-data.json
本课实际输入字段
| meeting | date | quote | topic |
|---|---|---|---|
| I1 | 2026-10-18 | The weight worries me. | weight |
| I2 | 2026-10-19 | The carton dimensions need checking. | dimensions |
| I3 | 2026-10-20 | The weight is acceptable; dimensions are still untested. | weight+dimensions |
保存Recipe提问
reference.csv：作者验收映射，不是目标API回包
| topic | earlier | current | sources |
|---|---|---|---|
| weight | worries | acceptable | ["I1", "I3"] |
| dimensions | untested | ["I2", "I3"] | |
| consensus | not claimed |
应当保留的约束
| topic | earlier | current | sources |
|---|---|---|---|
| weight | worries | acceptable | ["I1", "I3"] |
| dimensions | untested | ["I2", "I3"] | |
| consensus | not claimed |
故意构造的错误样例
| claim | fault |
|---|---|
| All concerns resolved | dimensions untested |
| Weight still rejected | ignores I3 |
先保存真实产品输出，再逐项对照。actualOutput / actualResult 在未执行时保持 not_run，不能把这里的期望值复制成已测结果。