Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →A股冰山订单/隐藏委托策略。当用户说"冰山订单"、"iceberg order"、"隐藏委托"、"冰山单"、"隐藏挂单"、"大单隐藏"时触发。基于 cn-stock-data 获取数据，分析冰山订单特征与检测方法。支持 formal/brief 两种输出风格。
.claude/skills/aifinlab-a-share-iceberg-order/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | 11% | 0% |
| case-16 | ✗→✓ | ▲ Improved | -36% | 0% |
| case-18 | ✗→✓ | ▲ Improved | -17% | 0% |
| case-19 | ✗→✓ | ▲ Improved | -20% | 0% |
| case-20 | ✗→✓ | ▲ Improved | -34% | 0% |
通过 cn-stock-data skill 获取数据：
# [标的] 冰山订单分析报告
## 一、检测结果
| 价位 | 方向 | 显示量 | 估计总量 | 置信度 |
|------|------|--------|---------|--------|
## 二、信号解读
[冰山单方向与含义]
## 三、策略建议
[跟随/回避建议]
## 四、监控要点## [标的] 冰山单速览
- 检测到买方冰山单@25.20
- 显示量500股，估计总量5万股
- 已持续30分钟，偏多信号
- 建议：25.20为强支撑，可轻仓跟随参考 references/iceberg-order-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 | 29,964 | 29,920 | -0% | 1 | 1 | 0% | 4,393 | 4,785 | +9% | 0 | 0 | — |
case-02 | fail→pass | 19,112 | 13,530 | -29% | 1 | 1 | 0% | 2,485 | 2,763 | +11% | 0 | 0 | — |
case-03 | fail→fail | 38,627 | 44,904 | +16% | 1 | 1 | 0% | 4,098 | 7,246 | +77% | 0 | 0 | — |
case-04 | pass→pass | 36,016 | 33,814 | -6% | 1 | 1 | 0% | 4,795 | 5,928 | +24% | 0 | 0 | — |
case-05 | pass→pass | 22,548 | 29,315 | +30% | 1 | 1 | 0% | 2,908 | 4,577 | +57% | 0 | 0 | — |
case-06 | pass→pass | 17,086 | 19,158 | +12% | 1 | 1 | 0% | 3,066 | 3,824 | +25% | 0 | 0 | — |
case-07 | pass→pass | 40,212 | 18,966 | -53% | 1 | 1 | 0% | 3,181 | 3,288 | +3% | 0 | 0 | — |
case-08 | pass→pass | 15,089 | 12,604 | -16% | 1 | 1 | 0% | 2,266 | 2,646 | +17% | 0 | 0 | — |
case-09 | pass→pass | 11,742 | 11,045 | -6% | 1 | 1 | 0% | 1,441 | 2,178 | +51% | 0 | 0 | — |
case-10 | pass→pass | 20,738 | 23,233 | +12% | 1 | 1 | 0% | 2,640 | 3,825 | +45% | 0 | 0 | — |
case-11 | pass→pass | 11,569 | 14,283 | +23% | 1 | 1 | 0% | 1,741 | 2,743 | +58% | 0 | 0 | — |
case-12 | pass→pass | 16,149 | 23,465 | +45% | 1 | 1 | 0% | 2,002 | 3,656 | +83% | 0 | 0 | — |
case-13 | pass→pass | 18,427 | 16,188 | -12% | 1 | 1 | 0% | 2,543 | 3,220 | +27% | 0 | 0 | — |
case-14 | pass→pass | 19,404 | 22,774 | +17% | 1 | 1 | 0% | 2,477 | 3,557 | +44% | 0 | 0 | — |
case-15 | pass→pass | 21,600 | 19,963 | -8% | 1 | 1 | 0% | 2,703 | 3,100 | +15% | 0 | 0 | — |
case-16 | fail→pass | 11,625 | 2,886 | -75% | 1 | 1 | 0% | 1,823 | 1,160 | -36% | 0 | 0 | — |
case-17 | pass→pass | 11,317 | 11,945 | +6% | 1 | 1 | 0% | 1,859 | 2,223 | +20% | 0 | 0 | — |
case-18 | fail→pass | 8,677 | 2,813 | -68% | 1 | 1 | 0% | 1,269 | 1,059 | -17% | 0 | 0 | — |
case-19 | fail→pass | 8,823 | 3,355 | -62% | 1 | 1 | 0% | 1,444 | 1,152 | -20% | 0 | 0 | — |
case-20 | fail→pass | 9,313 | 2,318 | -75% | 1 | 1 | 0% | 1,553 | 1,023 | -34% | 0 | 0 | — |
case-21 | fail→pass | 21,515 | 9,615 | -55% | 1 | 1 | 0% | 1,965 | 1,978 | +1% | 0 | 0 | — |
case-22 | pass→pass | 18,561 | 23,384 | +26% | 1 | 1 | 0% | 2,263 | 3,825 | +69% | 0 | 0 | — |
case-23 | pass→pass | 19,596 | 16,957 | -13% | 1 | 1 | 0% | 2,973 | 3,220 | +8% | 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. 23 cases were attempted. The headline lift of +26 percentage points is the difference between those two pass rates over the 23 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.