Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Read research outline, launch independent agent for each item for deep research. Disable task output.
.claude/skills/weizhena-research-deep/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | -3% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 17% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 28% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -16% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -6% | 0% |
/research-deep
在当前工作目录查找 */outline.yaml 文件,读取items列表、execution配置(含items_per_agent)。
参数获取:
{topic}: outline.yaml中的topic字段{item_name}: item的name字段{item_related_info}: item的完整yaml内容(name + category + description等){output_dir}: outline.yaml中execution.output_dir(默认./results){fields_path}: {topic}/fields.yaml的绝对路径{output_path}: {output_dir}/{item_name_slug}.json的绝对路径(slugify处理item_name:空格替换为_,移除特殊字符)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。
Prompt模板:
pythonprompt = f"""## 任务 调研 {item_related_info},输出结构化JSON到 {output_path} ## 字段定义 读取 {fields_path} 获取所有字段定义 ## 输出要求 1. 按fields.yaml定义的字段输出JSON 2. 不确定的字段值标注[不确定] 3. JSON末尾添加uncertain数组,列出所有不确定的字段名 4. 所有字段值必须使用中文输出(调研过程可用英文,但最终JSON值为中文) ## 输出路径 {output_path} ## 验证 完成JSON输出后,运行验证脚本确保字段完整覆盖: python ~/.claude/skills/research/validate_json.py -f {fields_path} -j {output_path} 验证通过后才算完成任务。 """
One-shot示例(假设调研GitHub Copilot):
## 任务
调研 name: GitHub Copilot
category: 国际产品
description: Microsoft/GitHub开发,首个主流AI编程助手,市场份额约40%,输出结构化JSON到 {project_dir}/results/GitHub_Copilot.json
## 字段定义
读取 {project_dir}/fields.yaml 获取所有字段定义
## 输出要求
1. 按fields.yaml定义的字段输出JSON
2. 不确定的字段值标注[不确定]
3. JSON末尾添加uncertain数组,列出所有不确定的字段名
4. 所有字段值必须使用中文输出(调研过程可用英文,但最终JSON值为中文)
## 输出路径
{project_dir}/results/GitHub_Copilot.json
## 验证
完成JSON输出后,运行验证脚本确保字段完整覆盖:
python ~/.claude/skills/research/validate_json.py -f {project_dir}/fields.yaml -j {project_dir}/results/GitHub_Copilot.json
验证通过后才算完成任务。全部完成后输出:
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 7,664 | 2,183 | -72% | 1 | 1 | 0% | 1,096 | 1,148 | +5% | 0 | 0 | — |
case-02 | fail→fail | 11,579 | 7,910 | -32% | 1 | 1 | 0% | 1,638 | 1,487 | -9% | 0 | 0 | — |
case-03 | fail→fail | 11,216 | 2,874 | -74% | 1 | 1 | 0% | 1,617 | 1,039 | -36% | 0 | 0 | — |
case-04 | fail→pass | 10,039 | 5,459 | -46% | 1 | 1 | 0% | 1,887 | 1,839 | -3% | 0 | 0 | — |
case-05 | fail→pass | 16,707 | 15,475 | -7% | 1 | 1 | 0% | 2,743 | 3,218 | +17% | 0 | 0 | — |
case-06 | fail→pass | 6,463 | 2,250 | -65% | 1 | 1 | 0% | 906 | 1,160 | +28% | 0 | 0 | — |
case-07 | fail→pass | 17,566 | 9,250 | -47% | 1 | 1 | 0% | 2,825 | 2,381 | -16% | 0 | 0 | — |
case-08 | fail→pass | 9,065 | 2,448 | -73% | 1 | 1 | 0% | 1,288 | 1,217 | -6% | 0 | 0 | — |
case-09 | pass→pass | 11,903 | 2,940 | -75% | 1 | 1 | 0% | 1,727 | 1,217 | -30% | 0 | 0 | — |
case-10 | fail→pass | 5,779 | 2,626 | -55% | 1 | 1 | 0% | 866 | 1,192 | +38% | 0 | 0 | — |
case-11 | pass→fail | 12,733 | 2,686 | -79% | 1 | 1 | 0% | 2,113 | 1,256 | -41% | 0 | 0 | — |
case-12 | fail→pass | 13,616 | 7,435 | -45% | 1 | 1 | 0% | 2,043 | 2,050 | +0% | 0 | 0 | — |
case-13 | pass→pass | 10,284 | 8,502 | -17% | 1 | 1 | 0% | 1,649 | 2,306 | +40% | 0 | 0 | — |
case-14 | fail→pass | 7,443 | 2,587 | -65% | 1 | 1 | 0% | 1,200 | 1,183 | -1% | 0 | 0 | — |
case-15 | fail→pass | 29,946 | 8,221 | -73% | 1 | 1 | 0% | 1,186 | 2,232 | +88% | 0 | 0 | — |
case-16 | fail→fail | 3,618 | 2,721 | -25% | 1 | 1 | 0% | 545 | 1,092 | +100% | 0 | 0 | — |
case-17 | fail→pass | 12,556 | 6,684 | -47% | 1 | 1 | 0% | 1,931 | 1,952 | +1% | 0 | 0 | — |
case-18 | fail→pass | 9,455 | 2,396 | -75% | 1 | 1 | 0% | 1,561 | 1,205 | -23% | 0 | 0 | — |
case-19 | fail→pass | 13,493 | 10,288 | -24% | 1 | 1 | 0% | 2,286 | 2,654 | +16% | 0 | 0 | — |
case-20 | fail→pass | 8,525 | 2,518 | -70% | 1 | 1 | 0% | 1,355 | 1,218 | -10% | 0 | 0 | — |
case-21 | fail→pass | 5,966 | 3,028 | -49% | 1 | 1 | 0% | 1,014 | 1,318 | +30% | 0 | 0 | — |
case-22 | fail→pass | 11,834 | 4,620 | -61% | 1 | 1 | 0% | 1,962 | 1,608 | -18% | 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 21 counted toward the lift figure. The other 1 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 +64 percentage points is the difference between those two pass rates over the 21 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.