Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Summarize deep research results into markdown report, cover all fields, skip uncertain values.
.claude/skills/weizhena-research-report/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-08 | ✗→✓ | ▲ Improved | -22% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -20% | 0% |
| case-12 | ✗→✓ | ▲ Improved | 8% | 0% |
| case-13 | ✗→✓ | ▲ Improved | 7% | 0% |
| case-20 | ✗→✓ | ▲ Improved | 60% | 0% |
/research-report
在当前工作目录查找 */outline.yaml,读取topic和output_dir配置。
读取所有JSON结果,提取适合在目录中显示的字段(数值型、简短指标),例如:
使用AskUserQuestion询问用户:
在 {topic}/ 目录下生成 generate_report.py,脚本要求:
{topic}/report.md目录格式要求:
1. [GitHub Copilot](#github-copilot) - Stars: 10k | Score: 85%1. JSON结构兼容 支持两种JSON结构:
{"name": "xxx", "release_date": "xxx"}{"basic_info": {"name": "xxx"}, "technical_features": {...}}字段查找顺序:顶层 -> category映射key -> 遍历所有嵌套dict
2. Category多语言映射 fields.yaml的category名与JSON的key可能是任意组合(中中、中英、英中、英英)。必须建立双向映射:
pythonCATEGORY_MAPPING = { "基本信息": ["basic_info", "基本信息"], "技术特性": ["technical_features", "technical_characteristics", "技术特性"], "性能指标": ["performance_metrics", "performance", "性能指标"], "里程碑意义": ["milestone_significance", "milestones", "里程碑意义"], "商业信息": ["business_info", "commercial_info", "商业信息"], "竞争与生态": ["competition_ecosystem", "competition", "竞争与生态"], "历史沿革": ["history", "历史沿革"], "市场定位": ["market_positioning", "market", "市场定位"], }
3. 复杂值格式化
| 分隔kv<br>或使用blockquote格式,提高可读性4. 额外字段收集 收集JSON中有但fields.yaml中没定义的字段,放入"其他信息"分类。注意过滤:
_source_file, uncertainbasic_info, technical_features等uncertain数组:需要逐行显示每个字段名,不要压缩成一行5. 不确定值跳过 跳过条件:
[不确定]字符串uncertain数组中运行 python {topic}/generate_report.py
{topic}/generate_report.py - 转换脚本{topic}/report.md - 汇总报告| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 3,388 | 3,275 | -3% | 1 | 1 | 0% | 245 | 1,167 | +376% | 0 | 0 | — |
case-02 | fail→fail | 17,973 | 4,997 | -72% | 1 | 1 | 0% | 3,903 | 1,359 | -65% | 0 | 0 | — |
case-03 | fail→fail | 28,189 | 3,150 | -89% | 1 | 1 | 0% | 5,604 | 1,073 | -81% | 0 | 0 | — |
case-04 | pass→pass | 17,635 | 19,261 | +9% | 1 | 1 | 0% | 3,476 | 4,736 | +36% | 0 | 0 | — |
case-05 | pass→pass | 8,046 | 12,527 | +56% | 1 | 1 | 0% | 1,434 | 3,541 | +147% | 0 | 0 | — |
case-06 | pass→pass | 22,283 | 19,998 | -10% | 1 | 1 | 0% | 5,032 | 5,364 | +7% | 0 | 0 | — |
case-07 | fail→fail | 10,543 | 9,339 | -11% | 1 | 1 | 0% | 1,863 | 2,704 | +45% | 0 | 0 | — |
case-08 | fail→pass | 12,212 | 4,980 | -59% | 1 | 1 | 0% | 2,358 | 1,847 | -22% | 0 | 0 | — |
case-09 | fail→fail | 11,414 | 7,129 | -38% | 1 | 1 | 0% | 2,111 | 2,455 | +16% | 0 | 0 | — |
case-10 | fail→pass | 9,121 | 2,222 | -76% | 1 | 1 | 0% | 1,617 | 1,297 | -20% | 0 | 0 | — |
case-11 | pass→pass | 10,789 | 8,574 | -21% | 1 | 1 | 0% | 2,148 | 2,438 | +14% | 0 | 0 | — |
case-12 | fail→pass | 9,593 | 5,853 | -39% | 1 | 1 | 0% | 1,936 | 2,092 | +8% | 0 | 0 | — |
case-13 | fail→pass | 9,649 | 4,992 | -48% | 1 | 1 | 0% | 1,700 | 1,818 | +7% | 0 | 0 | — |
case-14 | pass→pass | 11,290 | 4,287 | -62% | 1 | 1 | 0% | 1,951 | 1,760 | -10% | 0 | 0 | — |
case-15 | pass→pass | 12,319 | 13,198 | +7% | 1 | 1 | 0% | 2,449 | 3,905 | +59% | 0 | 0 | — |
case-16 | pass→pass | 9,676 | 8,242 | -15% | 1 | 1 | 0% | 1,751 | 2,607 | +49% | 0 | 0 | — |
case-17 | fail→fail | 8,032 | 2,078 | -74% | 1 | 1 | 0% | 1,499 | 1,330 | -11% | 0 | 0 | — |
case-18 | fail→fail | 6,806 | 2,023 | -70% | 1 | 1 | 0% | 1,339 | 1,255 | -6% | 0 | 0 | — |
case-19 | pass→pass | 12,460 | 6,134 | -51% | 1 | 1 | 0% | 2,324 | 2,027 | -13% | 0 | 0 | — |
case-20 | fail→pass | 9,245 | 9,190 | -1% | 1 | 1 | 0% | 1,642 | 2,627 | +60% | 0 | 0 | — |
case-21 | pass→pass | 11,525 | 5,366 | -53% | 1 | 1 | 0% | 1,791 | 1,898 | +6% | 0 | 0 | — |
case-22 | fail→pass | 11,692 | 6,693 | -43% | 1 | 1 | 0% | 2,249 | 2,071 | -8% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted, and 19 counted toward the lift figure. The other 3 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +27 percentage points is the difference between those two pass rates over the 19 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.