Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →当用户需要在银行风险管理场景下，围绕交易异常进行持续监测、异常识别和风险分级时使用本技能。适合输出异常信号摘要、优先级排序、排查建议与升级路径。
.claude/skills/aifinlab-bank-t203-risk-management-transaction-anomaly-assistant/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-17 | ✗→✓ | ▲ Improved | 55% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 26% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 25% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 49% | 0% |
| case-12 | ✗→✓ | ▲ Improved | 62% | 0% |
本技能面向银行风险管理场景，聚焦交易金额、频率、对手方、时间分布和渠道行为的异常识别。目标是将规则命中、交易趋势和客户画像变化转成可执行的排查动作与风险升级路径，支持风控、预警运营与合规排查团队。
scripts/transaction_anomaly_scan.py：用于交易异常规则命中、分级与摘要输出| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-02 | fail→fail | 27,675 | 36,398 | +32% | 1 | 1 | 0% | 3,355 | 4,045 | +21% | 0 | 0 | — |
case-16 | fail→fail | 21,777 | 27,562 | +27% | 1 | 1 | 0% | 2,885 | 4,441 | +54% | 0 | 0 | — |
case-17 | fail→pass | 17,966 | 32,440 | +81% | 1 | 1 | 0% | 2,265 | 3,513 | +55% | 0 | 0 | — |
case-01 | fail→fail | 25,094 | 37,838 | +51% | 1 | 1 | 0% | 3,118 | 3,883 | +25% | 0 | 0 | — |
case-03 | fail→fail | 14,007 | 20,690 | +48% | 1 | 1 | 0% | 1,899 | 3,339 | +76% | 0 | 0 | — |
case-04 | pass→pass | 10,027 | 15,455 | +54% | 1 | 1 | 0% | 1,137 | 2,734 | +140% | 0 | 0 | — |
case-05 | fail→fail | 25,291 | 39,215 | +55% | 1 | 1 | 0% | 4,919 | 6,569 | +34% | 0 | 0 | — |
case-06 | fail→pass | 23,127 | 23,690 | +2% | 1 | 1 | 0% | 3,389 | 4,263 | +26% | 0 | 0 | — |
case-07 | fail→fail | 19,565 | 18,955 | -3% | 1 | 1 | 0% | 2,874 | 3,396 | +18% | 0 | 0 | — |
case-08 | fail→pass | 20,570 | 23,222 | +13% | 1 | 1 | 0% | 2,965 | 3,718 | +25% | 0 | 0 | — |
case-09 | fail→pass | 22,209 | 23,348 | +5% | 1 | 1 | 0% | 2,483 | 3,699 | +49% | 0 | 0 | — |
case-10 | fail→fail | 23,223 | 21,101 | -9% | 1 | 1 | 0% | 3,157 | 3,784 | +20% | 0 | 0 | — |
case-11 | pass→pass | 18,314 | 23,360 | +28% | 1 | 1 | 0% | 2,541 | 3,486 | +37% | 0 | 0 | — |
case-12 | fail→pass | 16,062 | 20,122 | +25% | 1 | 1 | 0% | 2,272 | 3,670 | +62% | 0 | 0 | — |
case-13 | pass→pass | 19,901 | 22,427 | +13% | 1 | 1 | 0% | 2,899 | 3,636 | +25% | 0 | 0 | — |
case-14 | fail→pass | 23,516 | 24,996 | +6% | 1 | 1 | 0% | 3,171 | 4,205 | +33% | 0 | 0 | — |
case-15 | fail→pass | 19,800 | 20,045 | +1% | 1 | 1 | 0% | 2,512 | 3,649 | +45% | 0 | 0 | — |
case-18 | fail→pass | 22,880 | 24,452 | +7% | 1 | 1 | 0% | 3,022 | 3,773 | +25% | 0 | 0 | — |
case-19 | fail→fail | 23,574 | 23,028 | -2% | 1 | 1 | 0% | 3,313 | 3,835 | +16% | 0 | 0 | — |
case-20 | fail→pass | 45,436 | 23,427 | -48% | 1 | 1 | 0% | 2,817 | 3,909 | +39% | 0 | 0 | — |
case-21 | fail→pass | 22,859 | 17,105 | -25% | 1 | 1 | 0% | 1,972 | 2,868 | +45% | 0 | 0 | — |
case-22 | fail→pass | 23,897 | 27,911 | +17% | 1 | 1 | 0% | 3,156 | 4,301 | +36% | 0 | 0 | — |
case-23 | fail→fail | 28,147 | 20,908 | -26% | 1 | 1 | 0% | 3,537 | 3,706 | +5% | 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 +48 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.