Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Conduct preliminary research on a topic and generate research outline. For academic research, benchmark research, technology selection, etc.
.claude/skills/weizhena-research/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | 9% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 123% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 19% | 0% |
| case-12 | ✗→✓ | ▲ Improved | -10% | 0% |
| case-14 | ✗→✓ | ▲ Improved | 2% | 0% |
/research <topic>
基于topic,利用模型已有知识生成:
输出{step1_output},使用AskUserQuestion确认:
使用AskUserQuestion询问时间范围(如:最近6个月、2024年至今、不限)。
参数获取:
{topic}: 用户输入的调研话题{YYYY-MM-DD}: 当前日期{step1_output}: Step 1生成的完整输出内容{time_range}: 用户指定的时间范围硬约束:以下prompt必须严格复述,仅替换{xxx}中的变量,禁止改写结构或措辞。
启动1个web-search-agent(后台),Prompt模板:
pythonprompt = f"""## 任务 调研话题: {topic} 当前日期: {YYYY-MM-DD} 基于以下初步框架,补充最新items和推荐调研字段。 ## 已有框架 {step1_output} ## 目标 1. 验证已有items是否遗漏重要对象 2. 根据遗漏对象进行补充items 3. 继续搜索{topic}相关且{time_range}内的items并补充 4. 补充新fields ## 输出要求 直接返回结构化结果(不写文件): ### 补充Items - item_name: 简要说明(为什么应该加入) ... ### 推荐补充字段 - field_name: 字段描述(为什么需要这个维度) ... ### 信息来源 - [来源1](url1) - [来源2](url2) """
One-shot示例(假设调研AI Coding发展史):
## 任务
调研话题: AI Coding 发展史
当前日期: 2025-12-30
基于以下初步框架,补充最新items和推荐调研字段。
## 已有框架
### Items列表
1. GitHub Copilot: Microsoft/GitHub开发,首个主流AI编程助手
2. Cursor: AI-first IDE,基于VSCode
...
### 字段框架
- 基本信息: name, release_date, company
- 技术特性: underlying_model, context_window
...
## 目标
1. 验证已有items是否遗漏重要对象
2. 根据遗漏对象进行补充items
3. 继续搜索AI Coding 发展史相关且2024年至今内的items并补充
4. 补充新fields
## 输出要求
直接返回结构化结果(不写文件):
### 补充Items
- item_name: 简要说明(为什么应该加入)
...
### 推荐补充字段
- field_name: 字段描述(为什么需要这个维度)
...
### 信息来源
- [来源1](url1)
- [来源2](url2)使用AskUserQuestion询问用户是否有已定义的字段文件,如有则读取并合并。
合并{step1_output}、{step2_output}和用户已有字段,生成两个文件:
outline.yaml(items + 配置):
fields.yaml(字段定义):
./{topic_slug}/outline.yaml 和 fields.yaml{当前工作目录}/{topic_slug}/
├── outline.yaml # items列表 + execution配置
└── fields.yaml # 字段定义/research-add-items - 补充items/research-add-fields - 补充字段/research-deep - 开始深度调研| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 10,885 | 8,342 | -23% | 1 | 1 | 0% | 1,810 | 2,432 | +34% | 0 | 0 | — |
case-02 | fail→pass | 25,972 | 15,285 | -41% | 1 | 1 | 0% | 3,195 | 3,479 | +9% | 0 | 0 | — |
case-03 | fail→fail | 9,331 | 17,429 | +87% | 1 | 1 | 0% | 524 | 2,164 | +313% | 0 | 0 | — |
case-04 | fail→fail | 22,421 | 29,934 | +34% | 1 | 1 | 0% | 2,765 | 6,714 | +143% | 0 | 0 | — |
case-05 | fail→fail | 9,843 | 5,593 | -43% | 1 | 1 | 0% | 1,603 | 2,183 | +36% | 0 | 0 | — |
case-06 | fail→pass | 7,590 | 13,216 | +74% | 1 | 1 | 0% | 1,407 | 3,138 | +123% | 0 | 0 | — |
case-07 | fail→fail | 19,974 | 15,559 | -22% | 1 | 1 | 0% | 3,776 | 3,835 | +2% | 0 | 0 | — |
case-08 | fail→fail | 28,101 | 14,860 | -47% | 1 | 1 | 0% | 2,831 | 3,340 | +18% | 0 | 0 | — |
case-09 | fail→fail | 12,222 | 16,014 | +31% | 1 | 1 | 0% | 2,122 | 3,922 | +85% | 0 | 0 | — |
case-10 | fail→pass | 8,309 | 3,586 | -57% | 1 | 1 | 0% | 1,252 | 1,496 | +19% | 0 | 0 | — |
case-11 | pass→pass | 13,904 | 9,517 | -32% | 1 | 1 | 0% | 2,516 | 2,763 | +10% | 0 | 0 | — |
case-12 | fail→pass | 17,326 | 11,779 | -32% | 1 | 1 | 0% | 2,771 | 2,495 | -10% | 0 | 0 | — |
case-13 | pass→pass | 5,892 | 3,637 | -38% | 1 | 1 | 0% | 1,006 | 1,632 | +62% | 0 | 0 | — |
case-14 | fail→pass | 7,764 | 2,271 | -71% | 1 | 1 | 0% | 1,356 | 1,380 | +2% | 0 | 0 | — |
case-15 | fail→pass | 7,561 | 2,077 | -73% | 1 | 1 | 0% | 1,248 | 1,365 | +9% | 0 | 0 | — |
case-16 | fail→pass | 24,569 | 2,299 | -91% | 1 | 1 | 0% | 2,282 | 1,426 | -38% | 0 | 0 | — |
case-17 | fail→fail | 7,904 | 3,215 | -59% | 1 | 1 | 0% | 1,403 | 1,562 | +11% | 0 | 0 | — |
case-18 | pass→pass | 11,703 | 2,868 | -75% | 1 | 1 | 0% | 1,933 | 1,541 | -20% | 0 | 0 | — |
case-19 | pass→pass | 12,536 | 6,904 | -45% | 1 | 1 | 0% | 2,161 | 2,258 | +4% | 0 | 0 | — |
case-20 | pass→pass | 10,621 | 4,808 | -55% | 1 | 1 | 0% | 1,715 | 2,046 | +19% | 0 | 0 | — |
case-21 | fail→fail | 24,086 | 11,498 | -52% | 1 | 1 | 0% | 5,309 | 3,137 | -41% | 0 | 0 | — |
case-22 | fail→fail | 4,542 | 9,350 | +106% | 1 | 1 | 0% | 254 | 2,708 | +966% | 0 | 0 | — |
case-23 | fail→pass | 6,352 | 11,045 | +74% | 1 | 1 | 0% | 1,152 | 3,177 | +176% | 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. 23 cases were attempted, and 21 counted toward the lift figure. The other 2 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 +35 percentage points is the difference between those two pass rates over the 21 comparable cases.
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.