Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →A股ML市场状态聚类/无监督分类。当用户说"市场聚类"、"无监督分类"、"状态聚类"、"ML聚类"、"K-means市场"、"市场分类"、"聚类分析市场"时触发。基于 cn-stock-data 获取数据，使用无监督学习识别市场状态。支持 formal/brief 两种输出风格。
.claude/skills/aifinlab-a-share-ml-regime-cluster/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | 9% | 0% |
| case-11 | ✗→✓ | ▲ Improved | 23% | 0% |
| case-13 | ✗→✓ | ▲ Improved | 22% | 0% |
| case-20 | ✗→✓ | ▲ Improved | -17% | 0% |
| case-18 | ✓→✗ | ▼ Worse | 7% | 0% |
通过 cn-stock-data skill 获取数据：
# ML市场状态聚类报告
## 一、聚类结果
| 状态 | 特征 | 占比 | 当前概率 |
|------|------|------|----------|
## 二、状态转移
[转移概率矩阵]
## 三、当前状态
[判断结果与置信度]
## 四、策略建议
[当前状态下的最优策略]## 市场状态速览
- GMM识别4种状态
- 当前状态：震荡偏多(概率68%)
- 转向牛市概率 22%
- 建议：维持6成仓位，偏向动量因子参考 references/ml-regime-cluster-guide.md 获取详细方法论与 A股实证研究。
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 28,188 | 82,051 | +191% | 1 | 1 | 0% | 4,501 | 5,285 | +17% | 0 | 0 | — |
case-02 | fail→pass | 18,498 | 19,622 | +6% | 1 | 1 | 0% | 3,034 | 3,295 | +9% | 0 | 0 | — |
case-03 | fail→fail | 26,550 | 38,732 | +46% | 1 | 1 | 0% | 4,319 | 6,987 | +62% | 0 | 0 | — |
case-04 | pass→pass | 27,113 | 34,954 | +29% | 1 | 1 | 0% | 4,663 | 5,543 | +19% | 0 | 0 | — |
case-05 | pass→pass | 26,608 | 36,244 | +36% | 1 | 1 | 0% | 4,502 | 7,271 | +62% | 0 | 0 | — |
case-06 | pass→pass | 26,794 | 23,517 | -12% | 1 | 1 | 0% | 5,146 | 5,124 | -0% | 0 | 0 | — |
case-07 | pass→pass | 27,173 | 27,244 | +0% | 1 | 1 | 0% | 3,656 | 4,316 | +18% | 0 | 0 | — |
case-08 | pass→pass | 26,917 | 23,221 | -14% | 1 | 1 | 0% | 3,925 | 3,766 | -4% | 0 | 0 | — |
case-09 | pass→pass | 24,701 | 23,413 | -5% | 1 | 1 | 0% | 3,430 | 3,606 | +5% | 0 | 0 | — |
case-10 | pass→pass | 18,472 | 18,535 | +0% | 1 | 1 | 0% | 2,900 | 3,459 | +19% | 0 | 0 | — |
case-11 | fail→pass | 18,905 | 17,129 | -9% | 1 | 1 | 0% | 2,619 | 3,222 | +23% | 0 | 0 | — |
case-12 | pass→pass | 17,952 | 15,357 | -14% | 1 | 1 | 0% | 2,633 | 2,734 | +4% | 0 | 0 | — |
case-13 | fail→pass | 16,423 | 17,483 | +6% | 1 | 1 | 0% | 2,755 | 3,357 | +22% | 0 | 0 | — |
case-14 | pass→pass | 17,873 | 20,146 | +13% | 1 | 1 | 0% | 2,832 | 3,801 | +34% | 0 | 0 | — |
case-15 | pass→pass | 19,567 | 19,119 | -2% | 1 | 1 | 0% | 2,569 | 3,355 | +31% | 0 | 0 | — |
case-16 | pass→pass | 22,543 | 23,789 | +6% | 1 | 1 | 0% | 3,351 | 4,179 | +25% | 0 | 0 | — |
case-17 | pass→pass | 25,670 | 22,994 | -10% | 1 | 1 | 0% | 3,698 | 4,101 | +11% | 0 | 0 | — |
case-18 | pass→fail | 21,910 | 19,180 | -12% | 1 | 1 | 0% | 3,385 | 3,618 | +7% | 0 | 0 | — |
case-19 | fail→fail | 32,432 | 25,580 | -21% | 1 | 1 | 0% | 4,825 | 4,665 | -3% | 0 | 0 | — |
case-20 | fail→pass | 9,040 | 5,056 | -44% | 1 | 1 | 0% | 1,587 | 1,318 | -17% | 0 | 0 | — |
case-21 | pass→pass | 20,810 | 22,125 | +6% | 1 | 1 | 0% | 3,036 | 3,837 | +26% | 0 | 0 | — |
case-22 | pass→pass | 17,149 | 12,883 | -25% | 1 | 1 | 0% | 2,407 | 2,495 | +4% | 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 +14 percentage points is the difference between those two pass rates over the 22 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.
| Model | Method | Date | Lift |
|---|---|---|---|
| gemini-3.6-flash | verified | 8/28/2026 | +4% |
Other measured skills in the registry, with their headline benchmark lift.