Use when a complex problem needs a structured expert team rather than a single general answer. Runs a Single-CEO Expert Council with a Nuwa-style decision le...
Coding
Adaptive Expert Interview
Try itUse when conducting an in-depth adaptive interview with a domain expert to extract structured Q&A pairs. Triggers include need to capture expert knowledge, d...
What it does
Use when conducting an in-depth adaptive interview with a domain expert to extract structured Q&A pairs. Triggers include need to capture expert knowledge, decision frameworks, and cognitive boundaries through progressive questioning.
The skill document
自适应专家访谈
Overview
通过多轮对话对领域专家进行自适应深度访谈,自动生成从易到难的问题,挖掘决策风格和领域认知,输出标准化的问答对。
When to Use
- 需要从领域专家身上提取结构化知识
- 希望将隐性经验转化为显性原则
- 需要采集标准化的问答对供后续分析
Workflow
1. 初始化访谈
收集用户输入(领域、专家定位、目标数量、聚焦关键词),然后运行:
python interview_engine.py init \
--domain "领域名称" \
--expert "专家定位描述" \
--target 50 \
--keywords "关键词1,关键词2" \
--output ./interview_state.json
向用户展示预计时长和流程说明。
2. 获取状态上下文
每轮提问前,获取当前状态:
python interview_engine.py status --state ./interview_state.json
3. 生成问题
- 读取
prompts/generate_question.txt模板 - 将状态信息填充到模板变量中
- 将填充后的 Prompt 发送给 LLM
- LLM 返回 JSON 格式的问题
- 向用户展示问题
4. 获取回答
等待用户回答。
5. 分析回答
- 读取
prompts/analyze_response.txt模板 - 填充当前问题和用户回答
- 发送给 LLM 分析
- LLM 返回 JSON 格式的分析结果
6. 更新状态
运行:
python interview_engine.py update \
--state ./interview_state.json \
--q-id 1 \
--question "问题文本" \
--answer "回答文本" \
--analysis '{"dimension": "typical_cases", "needs_follow_up": false, "quality": "high"}'
引擎会自动:
- 更新进度
- 追踪话题覆盖
- 控制追问次数
- 检查阶段切换
- 判断是否进入快速模式
7. 循环
重复步骤 2-6,直到达到目标数量。
8. 导出结果
python interview_engine.py export \
--state ./interview_state.json \
--format jsonl \
--output ./interview_result.jsonl
中断恢复
如果访谈需要中断:
python interview_engine.py snapshot \
--state ./interview_state.json \
--output ./interview_backup.json
优先级规则
执行时遵循以下优先级:
- 不中断访谈 - 永远推进对话
- 完成目标数量 - 达到设定目标
- 控制时长 - 在承诺时间内完成
- 挖掘深度 - 时间允许时追问
输出格式
JSON Lines,每行一个对象:
{"q_id": 1, "phase": "warmup", "type": "choice", "dimension": "core_principles", "question": "...", "answer": "...", "timestamp": "..."}
Related skills
Extracts what the user actually wants instead of what they think they should want. Achieves this through one-question-at-a-time interview until ~95% confidence about the underlying intent. Use when an Use when 需要Development领域自动化处理、数据分析和流程编排时使用。不适用于无明确需求的模糊场景。
Distill decision contexts, options, trade-offs, and outcomes into structured decision records. Use when the user is facing a choice, has made a decision they...
Point at any topic to build a compounding, domain-general expertise knowledge base: a high-fidelity source corpus, track-record-weighted experts, reconciled...
Detect requests to find, hire, arrange, or escalate work to a qualified human when judgment, credentials, physical presence, accountability, or specialized e...
Turn interview notes into a structured candidate scorecard and hire recommendation. Use when asked to write an interview scorecard, a candidate evaluation, a...