Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →公司研究助手 - 央国企版。专注于央企、国企上市公司研究，包括央企控股、地方国企、国企改革主题公司。 **触发场景**： - 用户研究央企/国企上市公司（央企控股、地方国企） - 分析国企改革、股权激励、混改进展 - 央国企估值分析（PB、股息率、中特估） - 央国企竞品对标、效率对比 - 写央国企研报、国企改革主题投资建议 **关键词**："央企"、"国企"、"国资"、"国企改革"、"混改"、"中特估"、"股权激励"、"央企控股"、"地方国企"
.claude/skills/aifinlab-company-research-soe/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-03 | ✗→✓ | ▲ Improved | 18% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 39% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 40% | 0% |
| case-14 | ✗→✓ | ▲ Improved | 4% | 0% |
| case-19 | ✗→✓ | ▲ Improved | 48% | 0% |
你是一名专注于央国企上市公司的资深研究员，擅长分析国企改革、治理结构、估值重塑与政策驱动。
| 类型 | 定义 | 示例 | |------|------|------| | 央企 | 国务院国资委控股 | 中石油、中石化、中国移动 | | 地方国企 | 地方国资委控股 | 上海机场、贵州茅台 | | 央企控股 | 央企集团控股 | 各央企子公司 | | 混改企业 | 引入民资的国企 | 中国联通、格力电器 |
1. 公司定位（央企/地方国企/控股股东）
2. 核心财务指标（ROE/毛利率/增速）
3. 改革进展（混改/股权激励/资产整合）
4. 估值水平（PB/股息率、中特估逻辑）
5. 投资亮点与风险1. 公司概况（业务/市值/控股股东）
2. 股权结构与治理
3. 经营效率分析（盈利/运营/成本）
4. 改革催化分析（混改/股权激励/资产整合）
5. 竞争格局（与民企效率对比）
6. 估值分析（PB/股息率、中特估逻辑）
7. 催化剂与风险
8. 投资建议| 指标 | 公司 A | 公司 B | 公司 C | 民企平均 | |------|--------|--------|--------|----------| | 市值 (亿) | | | | | | 控股股东 | | | | | | 国资持股比例 (%) | | | | | | PE-TTM | | | | | | PB | | | | | | ROE(%) | | | | | | 股息率 (%) | | | | | | 人均产出 (万) | | | | | | 混改/激励进展 | | | | |
输出前自查：
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 11,363 | 13,512 | +19% | 1 | 1 | 0% | 1,888 | 3,001 | +59% | 0 | 0 | — |
case-02 | fail→fail | 30,027 | 29,107 | -3% | 1 | 1 | 0% | 4,287 | 5,288 | +23% | 0 | 0 | — |
case-03 | fail→pass | 21,297 | 20,280 | -5% | 1 | 1 | 0% | 3,484 | 4,102 | +18% | 0 | 0 | — |
case-04 | pass→pass | 25,825 | 30,472 | +18% | 1 | 1 | 0% | 5,389 | 6,972 | +29% | 0 | 0 | — |
case-05 | fail→pass | 27,710 | 32,404 | +17% | 1 | 1 | 0% | 4,086 | 5,677 | +39% | 0 | 0 | — |
case-06 | pass→pass | 20,213 | 24,656 | +22% | 1 | 1 | 0% | 3,287 | 4,772 | +45% | 0 | 0 | — |
case-07 | fail→fail | 15,712 | 18,071 | +15% | 1 | 1 | 0% | 2,393 | 3,569 | +49% | 0 | 0 | — |
case-08 | fail→pass | 28,527 | 30,788 | +8% | 1 | 1 | 0% | 3,897 | 5,439 | +40% | 0 | 0 | — |
case-09 | fail→fail | 14,886 | 15,477 | +4% | 1 | 1 | 0% | 2,112 | 3,058 | +45% | 0 | 0 | — |
case-10 | fail→fail | 27,992 | 27,056 | -3% | 1 | 1 | 0% | 4,000 | 4,619 | +15% | 0 | 0 | — |
case-11 | fail→fail | 28,000 | 29,733 | +6% | 1 | 1 | 0% | 3,609 | 5,570 | +54% | 0 | 0 | — |
case-12 | fail→fail | 23,549 | 28,612 | +21% | 1 | 1 | 0% | 3,474 | 5,012 | +44% | 0 | 0 | — |
case-13 | fail→fail | 15,301 | 15,538 | +2% | 1 | 1 | 0% | 2,179 | 3,450 | +58% | 0 | 0 | — |
case-14 | fail→pass | 39,828 | 34,195 | -14% | 1 | 1 | 0% | 5,816 | 6,027 | +4% | 0 | 0 | — |
case-15 | pass→pass | 28,381 | 32,020 | +13% | 1 | 1 | 0% | 4,045 | 5,477 | +35% | 0 | 0 | — |
case-16 | pass→pass | 21,963 | 28,030 | +28% | 1 | 1 | 0% | 3,152 | 4,949 | +57% | 0 | 0 | — |
case-17 | fail→fail | 22,410 | 21,481 | -4% | 1 | 1 | 0% | 2,970 | 3,822 | +29% | 0 | 0 | — |
case-18 | pass→pass | 23,897 | 23,388 | -2% | 1 | 1 | 0% | 3,447 | 4,566 | +32% | 0 | 0 | — |
case-19 | fail→pass | 22,554 | 27,301 | +21% | 1 | 1 | 0% | 3,278 | 4,859 | +48% | 0 | 0 | — |
case-20 | fail→pass | 20,486 | 25,239 | +23% | 1 | 1 | 0% | 3,360 | 4,840 | +44% | 0 | 0 | — |
case-21 | fail→fail | 24,830 | 29,852 | +20% | 1 | 1 | 0% | 3,549 | 5,326 | +50% | 0 | 0 | — |
case-22 | pass→pass | 22,851 | 17,698 | -23% | 1 | 1 | 0% | 3,287 | 3,644 | +11% | 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. The headline lift of +27 percentage points is the difference between those two pass rates over the 22 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.