Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →A股止损策略/风控规则量化。当用户说"止损"、"stop loss"、"止盈"、"风控"、"该不该割肉"、"设在哪"、"止损位"时触发。量化设计止损止盈策略。支持formal和brief风格。
.claude/skills/aifinlab-a-share-stop-loss/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-09 | ✗→✓ | ▲ Improved | -60% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -36% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -53% | 0% |
| case-17 | ✗→✓ | ▲ Improved | -14% | 0% |
| case-18 | ✗→✓ | ▲ Improved | 58% | 0% |
bashSCRIPTS="$SKILLS_ROOT/cn-stock-data/scripts" python "$SCRIPTS/cn_stock_data.py" kline --code [CODE] --freq daily --start [日期] python "$SCRIPTS/cn_stock_data.py" quote --code [CODE] python "$SCRIPTS/cn_stock_data.py" finance --code [CODE]
回测各止损方法在该股上的历史表现（避免的亏损 vs 误杀的盈利）
根据标的波动特征选择最优止损幅度
| 维度 | formal | brief | |------|--------|-------| | 止损方案 | 多方法对比 | 建议止损位 | | 回测结果 | 各方法胜率+收益 | 推荐方法 | | 止盈建议 | 止盈策略 | 目标价位 | 默认风格：brief。
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-09 | fail→pass | 15,662 | 2,784 | -82% | 1 | 1 | 0% | 2,290 | 927 | -60% | 0 | 0 | — |
case-01 | fail→fail | 31,368 | 9,521 | -70% | 1 | 1 | 0% | 3,801 | 900 | -76% | 0 | 0 | — |
case-02 | fail→fail | 30,426 | 8,796 | -71% | 1 | 1 | 0% | 4,198 | 964 | -77% | 0 | 0 | — |
case-03 | pass→pass | 12,345 | 22,072 | +79% | 1 | 1 | 0% | 1,720 | 2,718 | +58% | 0 | 0 | — |
case-04 | pass→fail | 10,058 | 10,844 | +8% | 1 | 1 | 0% | 1,702 | 1,058 | -38% | 0 | 0 | — |
case-10 | pass→fail | 15,177 | 10,266 | -32% | 1 | 1 | 0% | 2,436 | 1,012 | -58% | 0 | 0 | — |
case-05 | pass→fail | 14,914 | 24,537 | +65% | 1 | 1 | 0% | 2,285 | 1,116 | -51% | 0 | 0 | — |
case-06 | fail→fail | 10,430 | 10,890 | +4% | 1 | 1 | 0% | 1,747 | 1,236 | -29% | 0 | 0 | — |
case-07 | fail→pass | 10,760 | 3,404 | -68% | 1 | 1 | 0% | 1,826 | 1,163 | -36% | 0 | 0 | — |
case-08 | fail→pass | 22,253 | 5,712 | -74% | 1 | 1 | 0% | 2,160 | 1,025 | -53% | 0 | 0 | — |
case-11 | pass→pass | 15,108 | 9,915 | -34% | 1 | 1 | 0% | 2,250 | 1,948 | -13% | 0 | 0 | — |
case-12 | fail→fail | 15,295 | 9,492 | -38% | 1 | 1 | 0% | 2,208 | 1,164 | -47% | 0 | 0 | — |
case-13 | fail→fail | 23,271 | 12,836 | -45% | 1 | 1 | 0% | 3,110 | 1,550 | -50% | 0 | 0 | — |
case-14 | pass→fail | 14,710 | 7,290 | -50% | 1 | 1 | 0% | 2,620 | 917 | -65% | 0 | 0 | — |
case-15 | pass→pass | 16,317 | 14,798 | -9% | 1 | 1 | 0% | 2,442 | 3,026 | +24% | 0 | 0 | — |
case-16 | pass→pass | 8,431 | 4,645 | -45% | 1 | 1 | 0% | 1,061 | 1,049 | -1% | 0 | 0 | — |
case-17 | fail→pass | 7,660 | 3,014 | -61% | 1 | 1 | 0% | 1,115 | 958 | -14% | 0 | 0 | — |
case-18 | fail→pass | 10,589 | 10,323 | -3% | 1 | 1 | 0% | 1,407 | 2,223 | +58% | 0 | 0 | — |
case-19 | fail→fail | 1,615 | 2,922 | +81% | 1 | 1 | 0% | 219 | 987 | +351% | 0 | 0 | — |
case-20 | pass→fail | 41,406 | 9,355 | -77% | 1 | 1 | 0% | 5,451 | 1,048 | -81% | 0 | 0 | — |
case-21 | pass→fail | 28,057 | 10,349 | -63% | 1 | 1 | 0% | 4,313 | 1,026 | -76% | 0 | 0 | — |
case-22 | pass→pass | 31,387 | 24,991 | -20% | 1 | 1 | 0% | 5,629 | 5,335 | -5% | 0 | 0 | — |
case-23 | fail→pass | 8,927 | 6,383 | -28% | 1 | 1 | 0% | 1,555 | 1,610 | +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. 23 cases were attempted, and 12 counted toward the lift figure. The other 11 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 0 percentage points is the difference between those two pass rates over the 12 comparable cases. 8 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.
Other measured skills in the registry, with their headline benchmark lift.