synthesis-writer
将证据综合成基于“papers/extraction_table.csv”的结构化叙述(“output/SYNTHESIS.md”),包括局限性和偏见考虑。**触发**:综合、证据综合、系统评论写作、综合写作、SYNTHESIS.md。**使用时间**:系统评论完成筛选+提取(含偏见评估)后进入写作阶段(C4)。**如果**:还没有则跳过`papers/extraction_table.csv`(或协议/筛选尚未完成)。**网络**:无。**Guardrail**:以提取表为证据支撑;明显局限性与依赖偏倚;不要在无数据支撑时扩写结论。
📖 中文导读:将证据综合成基于“papers/extraction_table.csv”的结构化叙述(“output/SYNTHESIS.md”),包括局限性和偏见考虑。**触发**:综合、证据综合、系统评论写作、综合写作、SYNTHESIS.md。**使用时间**:系统评论完成筛选+提取(含偏见评估)后进入写作阶段(C4)。**如果**:还没有则跳过`papers/extraction_table.csv`(或协议/筛选尚未完成)。**网络**:无。**Guardrail**:以提取表为证据支撑;明显局限性与依赖偏倚;不要在无数据支撑时扩写结论。 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: synthesis-writer
description: |
Synthesize evidence into a structured narrative (output/SYNTHESIS.md) grounded in papers/extraction_table.csv, including limitations and bias considerations.
Trigger: synthesis, evidence synthesis, systematic review writing, 综合写作, SYNTHESIS.md.
Use when: systematic review 完成 screening+extraction(含 bias 评估)后进入写作阶段(C4)。
Skip if: 还没有 papers/extraction_table.csv(或 protocol/screening 尚未完成)。
Network: none.
Guardrail: 以 extraction table 为证据底座;明确局限性与偏倚;不要在无数据支撑时扩写结论。
Synthesis Writer (systematic review)
Goal: write a structured synthesis that is traceable back to extracted data.
Inputs
Required:
papers/extraction_table.csv
Optional:
DECISIONS.md(approval to write prose, if your process requires it)output/PROTOCOL.md(to restate scope and methods consistently)
Outputs
output/SYNTHESIS.md
Workflow
Check writing approval (if applicable)
- If your pipeline requires it, confirm
DECISIONS.mdindicates approval before writing prose.
- If your pipeline requires it, confirm
Describe the evidence base (methods snapshot)
- Summarize the included set using
papers/extraction_table.csv(counts, time window, study types). - Keep this strictly descriptive.
- Summarize the included set using
Theme-based synthesis
- Group studies by theme/intervention/outcome (based on extraction fields).
- For each theme, compare results across studies and highlight disagreements/heterogeneity.
Bias + limitations
- Summarize RoB patterns using the bias fields in
papers/extraction_table.csv. - Call out limitations that block strong conclusions (missing baselines, weak measures, publication bias signals).
- Summarize RoB patterns using the bias fields in
Conclusions (bounded)
- State only what the extracted evidence supports.
- Separate “supported conclusions” vs “needs more evidence”.
Suggested outline for output/SYNTHESIS.md
- Research questions + scope (from
output/PROTOCOL.md) - Methods (sources, screening, extraction)
- Included studies summary (table-driven)
- Findings by theme (table-driven)
- Risk of bias + limitations
- Implications + future work (bounded)
Definition of Done
- Every major claim in
output/SYNTHESIS.mdis traceable to specific fields/rows inpapers/extraction_table.csv. - Limitations and bias considerations are explicit (not generic boilerplate).
Troubleshooting
Issue: the synthesis starts inventing facts not in the table
Fix:
- Restrict claims to what is explicitly present in
papers/extraction_table.csv; move speculation to “needs more evidence”.
Issue: extraction table is too sparse to synthesize
Fix:
- Add missing extraction fields/values first (re-run
extraction-form/bias-assessor), then write.