Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →A股高频订单流/逐笔成交分析。当用户说"订单流"、"order flow"、"逐笔成交"、"主动买卖"、"成交明细"、"大单追踪"、"逐笔分析"时触发。基于 cn-stock-data 获取数据，分析订单流方向、大单行为、成交节奏。支持 formal/brief 两种输出风格。
.claude/skills/aifinlab-a-share-hft-order-flow/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | 3% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 16% | 0% |
| case-15 | ✗→✓ | ▲ Improved | 28% | 0% |
| case-04 | ✓→✓ | ＝ Same ✓ | -10% | 0% |
| case-06 | ✓→✓ | ＝ Same ✓ | 34% | 0% |
通过 cn-stock-data skill 获取数据：
# [标的] 订单流分析报告
## 一、订单流方向
| 时段 | 净买入(万) | 大单占比 | 方向 |
|------|-----------|---------|------|
## 二、大单行为
[大单统计、拆单识别]
## 三、成交节奏
[到达率分析、异常模式]
## 四、毒性评估
[VPIN、风控建议]## [标的] 订单流速览
- 净买入 +2,350万，主动买占比 58%
- 大单买卖比 1.3:1，大单偏多
- VPIN 0.18，毒性正常
- 尾盘订单流加速，关注明日开盘参考 references/hft-order-flow-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 | 37,003 | 56,136 | +52% | 1 | 1 | 0% | 5,532 | 8,746 | +58% | 0 | 0 | — |
case-02 | fail→pass | 16,074 | 11,459 | -29% | 1 | 1 | 0% | 2,367 | 2,433 | +3% | 0 | 0 | — |
case-03 | fail→fail | 21,134 | 25,026 | +18% | 1 | 1 | 0% | 3,266 | 4,498 | +38% | 0 | 0 | — |
case-04 | pass→pass | 26,133 | 16,396 | -37% | 1 | 1 | 0% | 3,685 | 3,327 | -10% | 0 | 0 | — |
case-05 | fail→pass | 22,263 | 17,603 | -21% | 1 | 1 | 0% | 2,970 | 3,444 | +16% | 0 | 0 | — |
case-06 | pass→pass | 18,965 | 19,008 | +0% | 1 | 1 | 0% | 3,112 | 4,169 | +34% | 0 | 0 | — |
case-07 | pass→pass | 32,428 | 27,876 | -14% | 1 | 1 | 0% | 4,263 | 4,648 | +9% | 0 | 0 | — |
case-08 | pass→pass | 15,326 | 16,593 | +8% | 1 | 1 | 0% | 2,523 | 3,312 | +31% | 0 | 0 | — |
case-09 | pass→pass | 26,359 | 22,330 | -15% | 1 | 1 | 0% | 3,768 | 3,938 | +5% | 0 | 0 | — |
case-10 | pass→pass | 19,108 | 17,442 | -9% | 1 | 1 | 0% | 2,721 | 3,303 | +21% | 0 | 0 | — |
case-11 | pass→pass | 19,307 | 12,937 | -33% | 1 | 1 | 0% | 2,798 | 2,759 | -1% | 0 | 0 | — |
case-12 | fail→fail | 15,980 | 9,878 | -38% | 1 | 1 | 0% | 2,133 | 1,011 | -53% | 0 | 0 | — |
case-13 | pass→pass | 21,207 | 20,880 | -2% | 1 | 1 | 0% | 3,037 | 3,611 | +19% | 0 | 0 | — |
case-14 | pass→pass | 20,710 | 16,655 | -20% | 1 | 1 | 0% | 3,324 | 3,427 | +3% | 0 | 0 | — |
case-15 | fail→pass | 23,035 | 25,235 | +10% | 1 | 1 | 0% | 3,009 | 3,848 | +28% | 0 | 0 | — |
case-16 | pass→pass | 14,156 | 19,010 | +34% | 1 | 1 | 0% | 2,029 | 3,599 | +77% | 0 | 0 | — |
case-17 | fail→fail | 13,163 | 12,271 | -7% | 1 | 1 | 0% | 1,918 | 1,362 | -29% | 0 | 0 | — |
case-18 | pass→pass | 26,126 | 13,615 | -48% | 1 | 1 | 0% | 3,267 | 2,836 | -13% | 0 | 0 | — |
case-19 | fail→fail | 17,015 | 19,711 | +16% | 1 | 1 | 0% | 2,485 | 3,259 | +31% | 0 | 0 | — |
case-20 | fail→fail | 30,511 | 32,186 | +5% | 1 | 1 | 0% | 5,881 | 7,473 | +27% | 0 | 0 | — |
case-21 | fail→fail | 21,077 | 45,090 | +114% | 1 | 1 | 0% | 4,306 | 7,894 | +83% | 0 | 0 | — |
case-22 | fail→fail | 26,870 | 24,710 | -8% | 1 | 1 | 0% | 3,592 | 4,280 | +19% | 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 20 counted toward the lift figure. The other 2 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 +14 percentage points is the difference between those two pass rates over the 20 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 | +5% |
Other measured skills in the registry, with their headline benchmark lift.