sn-research-synthesis

Research synthesis converts completed sub-reports into a structured judgment layer by extracting conclusion-relevant findings and producing 2-5 main-line judgments with evidence strength. It reads existing request and plan goals to generate a unified synthesis.md file.

Cross-dimensional analysis resolves conflicts by summarizing consensus, flagging key disagreements, and explaining why they occur due to timing, sample differences, or stakeholder stances. Uncertainty tracking preserves gaps explicitly and records conditions that could overturn conclusions.

Evidence grading assigns high, medium, or low confidence to each main judgment with justification and applicability conditions. This labeling ensures decision-makers understand the robustness of each finding extracted from the sub-reports.

Generating a synthesis.md requires reading request.md, plan.json, and all completed sub_reports/*.md files located in your report directory. These inputs provide the goals and scattered findings that the synthesis process converts into actionable judgments.

Research synthesis provides final-deck wiring guidance by offering structure and chapter allocation suggestions for the final report. It does not draft the reader-facing report itself, focusing instead on wiring the judgment layer to support downstream writing.

Research synthesis is not suited for workflows lacking multiple completed sub-reports covering different dimensions. It requires existing sub-findings to extract conclusion-relevant threads and cannot generate judgments without prior dimensional research outputs.