Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →A股信用风险量化/违约概率分析。当用户说"信用风险"、"违约概率"、"credit risk"、"PD"、"信用评分"、"违约预警"、"信用量化"时触发。基于 cn-stock-data 获取数据，量化评估上市公司信用风险。支持 formal/brief 两种输出风格。
.claude/skills/aifinlab-a-share-credit-risk-quant/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-07 | ✗→✓ | ▲ Improved | 3% | 0% |
| case-15 | ✗→✓ | ▲ Improved | -20% | 0% |
| case-16 | ✗→✓ | ▲ Improved | -25% | 0% |
| case-17 | ✗→✓ | ▲ Improved | 2% | 0% |
| case-19 | ✗→✓ | ▲ Improved | -30% | 0% |
通过 cn-stock-data skill 获取数据：
# [标的] 信用风险量化报告
## 一、违约概率
| 模型 | PD | 评级 |
|------|-----|------|
## 二、财务健康
[Z-Score、关键财务指标]
## 三、趋势分析
[PD时序变化]
## 四、风险提示## [标的] 信用风险速览
- Merton PD = 0.8%，信用良好
- Z-Score = 2.5，灰色区域
- 资产负债率 55%，中等
- 建议：关注现金流变化趋势参考 references/credit-risk-quant-guide.md 获取详细方法论与 A股实证研究。
python# 调用 skill result = run_skill({ "param1": "value1", "param2": "value2" })
bashpython scripts/run_skill.py --input data.json
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 31,254 | 57,090 | +83% | 1 | 1 | 0% | 5,056 | 8,569 | +69% | 0 | 0 | — |
case-02 | fail→fail | 14,962 | 10,519 | -30% | 1 | 1 | 0% | 2,214 | 1,323 | -40% | 0 | 0 | — |
case-03 | pass→fail | 25,224 | 42,887 | +70% | 1 | 1 | 0% | 4,495 | 6,169 | +37% | 0 | 0 | — |
case-04 | pass→pass | 19,491 | 23,360 | +20% | 1 | 1 | 0% | 3,483 | 4,997 | +43% | 0 | 0 | — |
case-05 | pass→pass | 15,938 | 14,990 | -6% | 1 | 1 | 0% | 2,823 | 3,419 | +21% | 0 | 0 | — |
case-06 | pass→pass | 22,906 | 25,484 | +11% | 1 | 1 | 0% | 3,901 | 4,531 | +16% | 0 | 0 | — |
case-07 | fail→pass | 20,087 | 16,352 | -19% | 1 | 1 | 0% | 3,088 | 3,175 | +3% | 0 | 0 | — |
case-08 | pass→pass | 20,107 | 19,672 | -2% | 1 | 1 | 0% | 3,052 | 3,610 | +18% | 0 | 0 | — |
case-09 | fail→fail | 28,228 | 22,079 | -22% | 1 | 1 | 0% | 4,089 | 1,768 | -57% | 0 | 0 | — |
case-10 | fail→fail | 12,249 | 9,570 | -22% | 1 | 1 | 0% | 1,792 | 1,400 | -22% | 0 | 0 | — |
case-11 | pass→pass | 16,894 | 11,009 | -35% | 1 | 1 | 0% | 2,911 | 1,640 | -44% | 0 | 0 | — |
case-12 | pass→pass | 24,432 | 23,518 | -4% | 1 | 1 | 0% | 3,069 | 4,076 | +33% | 0 | 0 | — |
case-13 | pass→pass | 11,249 | 12,246 | +9% | 1 | 1 | 0% | 1,967 | 2,668 | +36% | 0 | 0 | — |
case-14 | pass→pass | 10,287 | 5,924 | -42% | 1 | 1 | 0% | 1,720 | 1,695 | -1% | 0 | 0 | — |
case-15 | fail→pass | 33,545 | 18,763 | -44% | 1 | 1 | 0% | 5,006 | 4,011 | -20% | 0 | 0 | — |
case-16 | fail→pass | 13,012 | 5,054 | -61% | 1 | 1 | 0% | 1,929 | 1,454 | -25% | 0 | 0 | — |
case-17 | fail→pass | 16,675 | 12,341 | -26% | 1 | 1 | 0% | 2,388 | 2,435 | +2% | 0 | 0 | — |
case-18 | pass→pass | 24,234 | 21,711 | -10% | 1 | 1 | 0% | 3,524 | 3,976 | +13% | 0 | 0 | — |
case-19 | fail→pass | 16,978 | 7,750 | -54% | 1 | 1 | 0% | 2,747 | 1,927 | -30% | 0 | 0 | — |
case-20 | pass→pass | 8,152 | 7,290 | -11% | 1 | 1 | 0% | 1,183 | 2,047 | +73% | 0 | 0 | — |
case-21 | fail→pass | 7,258 | 4,406 | -39% | 1 | 1 | 0% | 1,124 | 1,441 | +28% | 0 | 0 | — |
case-22 | pass→pass | 10,749 | 11,364 | +6% | 1 | 1 | 0% | 1,787 | 2,499 | +40% | 0 | 0 | — |
case-23 | fail→fail | 31,598 | 10,511 | -67% | 1 | 1 | 0% | 5,417 | 1,141 | -79% | 0 | 0 | — |
case-24 | pass→pass | 33,840 | 30,639 | -9% | 1 | 1 | 0% | 4,518 | 5,146 | +14% | 0 | 0 | — |
case-25 | pass→pass | 20,934 | 27,891 | +33% | 1 | 1 | 0% | 3,213 | 4,544 | +41% | 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. 25 cases were attempted, and 21 counted toward the lift figure. The other 4 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 +20 percentage points is the difference between those two pass rates over the 21 comparable cases. 3 cases got worse with the skill loaded, and they are 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 | +26% |
Other measured skills in the registry, with their headline benchmark lift.