Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →A股高频波动率/已实现方差分析。当用户说"已实现方差"、"realized variance"、"高频波动率估计"、"RV估计"、"二次变差"、"积分波动率"时触发。基于 cn-stock-data 获取数据，进行高频波动率估计与分析。支持 formal/brief 两种输出风格。
.claude/skills/aifinlab-a-share-realized-variance/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-16 | ✗→✓ | ▲ Improved | -51% | 0% |
| case-04 | ✓→✓ | ＝ Same ✓ | 4% | 0% |
| case-05 | ✓→✓ | ＝ Same ✓ | 15% | 0% |
| case-06 | ✓→✓ | ＝ Same ✓ | 34% | 0% |
| case-07 | ✓→✓ | ＝ Same ✓ | 37% | 0% |
通过 cn-stock-data skill 获取数据：
# [标的] 已实现方差分析报告
## 一、RV估计
| 指标 | 数值 | 年化 |
|------|------|------|
| 日RV | 0.032% | 25.3% |
## 二、跳跃分析
[跳跃检测、跳跃占比]
## 三、波动率预测
[HAR-RV预测结果]
## 四、方差风险溢价
[VRP水平与交易信号]## [标的] 已实现方差速览
- 日RV 0.032% (年化25.3%)
- 跳跃占比 18%，今日检测到1次跳跃
- HAR-RV预测明日RV：0.028%
- VRP +3.2%，卖波动率有溢价参考 references/realized-variance-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 | 40,907 | 48,084 | +18% | 1 | 1 | 0% | 5,963 | 7,428 | +25% | 0 | 0 | — |
case-02 | fail→fail | 12,821 | 11,953 | -7% | 1 | 1 | 0% | 2,135 | 1,319 | -38% | 0 | 0 | — |
case-03 | fail→fail | 40,127 | 28,338 | -29% | 1 | 1 | 0% | 5,779 | 6,741 | +17% | 0 | 0 | — |
case-04 | pass→pass | 21,043 | 17,372 | -17% | 1 | 1 | 0% | 3,400 | 3,531 | +4% | 0 | 0 | — |
case-05 | pass→pass | 20,662 | 19,527 | -5% | 1 | 1 | 0% | 3,982 | 4,595 | +15% | 0 | 0 | — |
case-06 | pass→pass | 16,629 | 17,849 | +7% | 1 | 1 | 0% | 2,836 | 3,806 | +34% | 0 | 0 | — |
case-07 | pass→pass | 19,239 | 20,786 | +8% | 1 | 1 | 0% | 3,209 | 4,383 | +37% | 0 | 0 | — |
case-08 | pass→pass | 18,760 | 19,773 | +5% | 1 | 1 | 0% | 3,139 | 4,112 | +31% | 0 | 0 | — |
case-09 | pass→pass | 20,165 | 17,635 | -13% | 1 | 1 | 0% | 2,492 | 3,294 | +32% | 0 | 0 | — |
case-10 | pass→pass | 20,312 | 26,401 | +30% | 1 | 1 | 0% | 2,942 | 4,506 | +53% | 0 | 0 | — |
case-11 | pass→pass | 18,169 | 18,404 | +1% | 1 | 1 | 0% | 2,846 | 3,583 | +26% | 0 | 0 | — |
case-12 | pass→pass | 18,604 | 17,395 | -6% | 1 | 1 | 0% | 3,232 | 3,876 | +20% | 0 | 0 | — |
case-13 | pass→pass | 20,627 | 16,427 | -20% | 1 | 1 | 0% | 3,240 | 3,416 | +5% | 0 | 0 | — |
case-14 | pass→pass | 22,211 | 24,461 | +10% | 1 | 1 | 0% | 3,528 | 4,289 | +22% | 0 | 0 | — |
case-15 | pass→pass | 18,295 | 15,519 | -15% | 1 | 1 | 0% | 2,770 | 3,074 | +11% | 0 | 0 | — |
case-16 | fail→pass | 20,474 | 5,290 | -74% | 1 | 1 | 0% | 3,159 | 1,542 | -51% | 0 | 0 | — |
case-17 | fail→fail | 11,245 | 12,871 | +14% | 1 | 1 | 0% | 1,912 | 2,702 | +41% | 0 | 0 | — |
case-18 | fail→fail | 37,080 | 35,548 | -4% | 1 | 1 | 0% | 5,767 | 7,173 | +24% | 0 | 0 | — |
case-19 | pass→pass | 19,944 | 22,998 | +15% | 1 | 1 | 0% | 3,048 | 3,897 | +28% | 0 | 0 | — |
case-20 | pass→pass | 17,907 | 20,751 | +16% | 1 | 1 | 0% | 4,123 | 5,367 | +30% | 0 | 0 | — |
case-21 | pass→pass | 20,038 | 22,225 | +11% | 1 | 1 | 0% | 3,205 | 4,452 | +39% | 0 | 0 | — |
case-22 | pass→pass | 21,323 | 28,206 | +32% | 1 | 1 | 0% | 3,987 | 4,952 | +24% | 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 +5 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.
| Model | Method | Date | Lift |
|---|---|---|---|
| gemini-3.6-flash | verified | 8/28/2026 | +17% |
Other measured skills in the registry, with their headline benchmark lift.