Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →A股尾部风险/黑天鹅/极端风险分析。当用户说"尾部风险"、"tail risk"、"黑天鹅"、"极端风险"、"肥尾"、"千股跌停"时触发。量化分析极端市场风险。支持formal和brief风格。
.claude/skills/aifinlab-a-share-tail-risk/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | 15% | 0% |
| case-17 | ✗→✓ | ▲ Improved | -41% | 0% |
| case-06 | ✓→✗ | ▼ Worse | -72% | 0% |
| case-08 | ✓→✗ | ▼ Worse | -66% | 0% |
| case-09 | ✓→✗ | ▼ Worse | -65% | 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]
| 维度 | formal | brief | |------|--------|-------| | 分布特征 | 偏度/峰度/QQ图 | 肥尾程度 | | 极端事件 | 历史事件详细 | 近期风险 | | 保护建议 | 对冲方案 | 风险等级 | 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-01 | fail→fail | 25,728 | 17,959 | -30% | 1 | 1 | 0% | 3,831 | 1,074 | -72% | 0 | 0 | — |
case-02 | fail→pass | 35,282 | 88,604 | +151% | 1 | 1 | 0% | 5,311 | 6,124 | +15% | 0 | 0 | — |
case-03 | fail→fail | 24,480 | 15,696 | -36% | 1 | 1 | 0% | 3,766 | 1,257 | -67% | 0 | 0 | — |
case-04 | fail→fail | 14,987 | 17,208 | +15% | 1 | 1 | 0% | 2,372 | 1,080 | -54% | 0 | 0 | — |
case-05 | fail→fail | 28,219 | 14,546 | -48% | 1 | 1 | 0% | 4,487 | 1,149 | -74% | 0 | 0 | — |
case-06 | pass→fail | 31,777 | 39,704 | +25% | 1 | 1 | 0% | 3,947 | 1,117 | -72% | 0 | 0 | — |
case-07 | pass→pass | 19,060 | 22,243 | +17% | 1 | 1 | 0% | 2,763 | 3,908 | +41% | 0 | 0 | — |
case-08 | pass→fail | 44,015 | 20,580 | -53% | 1 | 1 | 0% | 7,093 | 2,393 | -66% | 0 | 0 | — |
case-09 | pass→fail | 27,221 | 12,528 | -54% | 1 | 1 | 0% | 3,928 | 1,385 | -65% | 0 | 0 | — |
case-10 | pass→fail | 24,402 | 12,166 | -50% | 1 | 1 | 0% | 3,257 | 1,348 | -59% | 0 | 0 | — |
case-11 | fail→fail | 20,707 | 9,243 | -55% | 1 | 1 | 0% | 3,559 | 1,291 | -64% | 0 | 0 | — |
case-12 | fail→fail | 23,592 | 7,755 | -67% | 1 | 1 | 0% | 3,676 | 991 | -73% | 0 | 0 | — |
case-13 | pass→fail | 25,309 | 9,543 | -62% | 1 | 1 | 0% | 3,574 | 907 | -75% | 0 | 0 | — |
case-14 | fail→fail | 19,102 | 28,146 | +47% | 1 | 1 | 0% | 3,536 | 867 | -75% | 0 | 0 | — |
case-15 | pass→fail | 11,044 | 11,871 | +7% | 1 | 1 | 0% | 1,556 | 1,005 | -35% | 0 | 0 | — |
case-16 | pass→pass | 28,613 | 32,875 | +15% | 1 | 1 | 0% | 5,196 | 4,403 | -15% | 0 | 0 | — |
case-17 | fail→pass | 15,380 | 6,493 | -58% | 1 | 1 | 0% | 2,639 | 1,551 | -41% | 0 | 0 | — |
case-18 | pass→pass | 24,161 | 22,348 | -8% | 1 | 1 | 0% | 3,345 | 3,404 | +2% | 0 | 0 | — |
case-19 | fail→fail | 26,392 | 9,405 | -64% | 1 | 1 | 0% | 3,532 | 1,221 | -65% | 0 | 0 | — |
case-20 | pass→fail | 29,496 | 9,653 | -67% | 1 | 1 | 0% | 5,564 | 1,080 | -81% | 0 | 0 | — |
case-21 | pass→fail | 22,057 | 13,532 | -39% | 1 | 1 | 0% | 4,207 | 2,006 | -52% | 0 | 0 | — |
case-22 | pass→pass | 24,803 | 27,219 | +10% | 1 | 1 | 0% | 3,303 | 3,806 | +15% | 0 | 0 | — |
case-23 | fail→fail | 23,716 | 10,114 | -57% | 1 | 1 | 0% | 3,943 | 1,033 | -74% | 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 6 counted toward the lift figure. The other 17 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 -26 percentage points is the difference between those two pass rates over the 6 comparable cases. 13 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.