extraction-form
使用协议的提取模式将研究数据提取到结构化表(`papers/extraction_table.csv`)中。**触发器**:提取表格、提取表、数据提取、信息提取、提取表。**使用时间**:系统审查在筛选后进入提取(C3),需要把导入论文按字段落到CSV以支持后续合成。**如果**跳过:还没有`papers/screening_log.csv`或协议未锁定。**网络**: none.**Guardrail**:严格按schema填字段;不要在这个阶段写叙事综合(那是`synthesis-writer`)。
📖 中文导读:使用协议的提取模式将研究数据提取到结构化表(`papers/extraction_table.csv`)中。**触发器**:提取表格、提取表、数据提取、信息提取、提取表。**使用时间**:系统审查在筛选后进入提取(C3),需要把导入论文按字段落到CSV以支持后续合成。**如果**跳过:还没有`papers/screening_log.csv`或协议未锁定。**网络**: none.**Guardrail**:严格按schema填字段;不要在这个阶段写叙事综合(那是`synthesis-writer`)。 以下正文为该技能的原始 SKILL.md 内容(英文),上方安装命令可直接使用。
name: extraction-form
description: |
Extract study data into a structured table (papers/extraction_table.csv) using the protocol’s extraction schema.
Trigger: extraction form, extraction table, data extraction, 信息提取, 提取表.
Use when: systematic review 在 screening 后进入 extraction(C3),需要把纳入论文按字段落到 CSV 以支持后续 synthesis。
Skip if: 还没有 papers/screening_log.csv 或 protocol 未锁定。
Network: none.
Guardrail: 严格按 schema 填字段;不要在此阶段写 narrative synthesis(那是 synthesis-writer)。
Extraction Form (systematic review)
Goal: create a consistent, analysis-ready extraction table that is directly grounded in the protocol.
Inputs
Required:
papers/screening_log.csvoutput/PROTOCOL.md
Optional:
papers/paper_notes.jsonl(if you already have structured notes)
Outputs
papers/extraction_table.csv
Workflow
Determine the included set
- From
papers/screening_log.csv, collect all rows withdecision=include.
- From
Build/confirm the schema
- Use the extraction schema defined in
output/PROTOCOL.md. - If the protocol does not define fields yet, stop and update
output/PROTOCOL.mdfirst.
- Use the extraction schema defined in
Populate
papers/extraction_table.csv- One row per included paper.
- If
papers/paper_notes.jsonlexists, use it as a structured source for values/provenance (but keep the table schema governed byoutput/PROTOCOL.md). - Always include provenance columns:
paper_id,title,year,url
- For each protocol-defined field:
- fill concrete values (units explicit)
- use an explicit sentinel for unknowns (recommended: empty cell +
notes)
Keep it auditable
- If a value is inferred (not directly stated), mark it in a notes column.
- Do not write synthesis; only extraction.
Quick QA
- Ensure 1:1 coverage: included papers == extraction rows.
- Spot-check a few rows against the paper text/notes.
Definition of Done
-
papers/extraction_table.csvexists. - Every included paper from
papers/screening_log.csvhas exactly one extraction row. - Column meanings match
output/PROTOCOL.md(no ad-hoc columns without updating the protocol).
Troubleshooting
Issue: the protocol does not specify extraction fields
Fix:
- Update
output/PROTOCOL.md(extraction schema section) and re-run extraction.
Issue: extraction table mixes narrative text with fields
Fix:
- Move narrative into a
notescolumn and keep the rest as atomic values (numbers/enums/short strings).