Install
openclaw skills install @billb13/expert-distillerUse 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.
openclaw skills install @billb13/expert-distiller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。
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。
每轮提问前,获取当前状态:
python interview_engine.py status --state ./interview_state.json
prompts/generate_question.txt 模板等待用户回答。
prompts/analyze_response.txt 模板运行:
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:
重复步骤 2-6,直到达到目标数量。
python interview_engine.py export \
--state ./interview_state.json \
--format jsonl \
--output ./interview_result.jsonl
如果访谈需要中断:
python interview_engine.py snapshot \
--state ./interview_state.json \
--output ./interview_backup.json
执行时遵循以下优先级:
JSON Lines,每行一个对象:
{"q_id": 1, "phase": "warmup", "type": "choice", "dimension": "core_principles", "question": "...", "answer": "...", "timestamp": "..."}