
This is the final assembly step for research reports. It takes synthesis files, sub-reports, and drafts from deep research workflows and turns them into a structured markdown document with tables, Mermaid diagrams, and AI-generated concept images. The skill won't do new research or web searches. It's for when you already have the materials and need them packaged for an actual reader. The visual planning is surprisingly detailed, with rules for when to use flowcharts versus tables, how to color-code Mermaid nodes by semantic role, and when AI images make sense versus when they obscure precision. If you just ask for "research on X" without existing files, it routes you to sn-deep-research instead.
$ npx -y skills add opensensenova/sensenova-skills --skill sn-research-report --agent claude-codeInstalls into .claude/skills of the current project.
报告结构的决定因素是读者需要完成的认知任务,不是研究的信息领域。同一批信息,为做决策而读和为建立全貌而读,需要完全不同的组织方式。
digraph select {
"研究任务" [shape=doublecircle];
"领域专属?" [shape=diamond];
"主认知任务?" [shape=diamond];
"学术综述" [shape=box];
"医疗健康" [shape=box];
"法律研究" [shape=box];
"政策分析" [shape=box];
"全景叙述" [shape=box];
"对比选型" [shape=box];
"实体调查" [shape=box];
"时序追踪" [shape=box];
"研究任务" -> "领域专属?";
"领域专属?" -> "学术综述" [label="学术综述"];
"领域专属?" -> "医疗健康" [label="医疗方案报告"];
"领域专属?" -> "法律研究" [label="法律备忘录"];
"领域专属?" -> "政策分析" [label="政策简报"];
"领域专属?" -> "主认知任务?" [label="否"];
"主认知任务?" -> "全景叙述" [label="了解全貌"];
"主认知任务?" -> "对比选型" [label="比较/选择"];
"主认知任务?" -> "实体调查" [label="深挖某对象"];
"主认知任务?" -> "时序追踪" [label="还原事件"];
}
领域专属触发条件:最终产出是学术综述、医疗方案报告、法律备忘录或政策简报,且读者对该类文档有强烈格式预期。商业/投资场景下涉及医疗或法律内容,优先按认知任务选模板,再叠加领域术语惯例。
复合意图处理:主意图决定整体框架,次意图压缩为一个章节。例如"先了解行业再评估某公司"→主意图是实体调查(论点驱动式),全景内容压缩为"行业背景"章节,不重复建立完整全景框架。
适用:行业研究、可行性研究、技术趋势综述
读者认知任务:建立对一个空间的完整心智地图
Sponsors