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，对零售客户可疑交易进行监测、分级与排查建议输出时使用本技能。适合生成命中说明、优先级判断、排查清单与升级路径。
.claude/skills/aifinlab-bank-t205-risk-management-aml-identification-assistant/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-03 | ✗→✓ | ▲ Improved | 14% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 30% | 0% |
| case-11 | ✗→✓ | ▲ Improved | -2% | 0% |
| case-14 | ✗→✓ | ▲ Improved | 21% | 0% |
| case-07 | ✓→✓ | ＝ Same ✓ | 36% | 0% |
本技能面向银行风险管理场景，聚焦零售客户的可疑交易识别与反洗钱排查支持。目标是将规则命中、交易行为与客户画像变化形成可执行的排查清单和升级路径，帮助反洗钱、合规与预警运营团队快速响应。
scripts/aml_retail_screening.py：用于零售可疑交易规则命中、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-07 | pass→pass | 18,479 | 25,995 | +41% | 1 | 1 | 0% | 2,416 | 3,275 | +36% | 0 | 0 | — |
case-01 | fail→fail | 20,003 | 33,611 | +68% | 1 | 1 | 0% | 3,298 | 4,176 | +27% | 0 | 0 | — |
case-02 | fail→fail | 27,929 | 21,002 | -25% | 1 | 1 | 0% | 3,609 | 4,013 | +11% | 0 | 0 | — |
case-03 | fail→pass | 22,491 | 23,624 | +5% | 1 | 1 | 0% | 3,456 | 3,954 | +14% | 0 | 0 | — |
case-04 | pass→pass | 17,178 | 21,820 | +27% | 1 | 1 | 0% | 2,439 | 3,352 | +37% | 0 | 0 | — |
case-05 | fail→fail | 25,222 | 25,392 | +1% | 1 | 1 | 0% | 3,939 | 4,667 | +18% | 0 | 0 | — |
case-06 | pass→pass | 20,303 | 20,310 | +0% | 1 | 1 | 0% | 2,525 | 3,248 | +29% | 0 | 0 | — |
case-08 | pass→pass | 24,543 | 19,533 | -20% | 1 | 1 | 0% | 2,708 | 3,572 | +32% | 0 | 0 | — |
case-09 | fail→pass | 22,418 | 23,639 | +5% | 1 | 1 | 0% | 2,862 | 3,716 | +30% | 0 | 0 | — |
case-10 | pass→pass | 21,741 | 30,205 | +39% | 1 | 1 | 0% | 2,864 | 3,435 | +20% | 0 | 0 | — |
case-11 | fail→pass | 18,875 | 16,256 | -14% | 1 | 1 | 0% | 2,984 | 2,922 | -2% | 0 | 0 | — |
case-12 | pass→pass | 18,476 | 20,172 | +9% | 1 | 1 | 0% | 2,634 | 3,341 | +27% | 0 | 0 | — |
case-13 | pass→pass | 11,649 | 15,175 | +30% | 1 | 1 | 0% | 1,730 | 2,625 | +52% | 0 | 0 | — |
case-14 | fail→pass | 21,631 | 21,320 | -1% | 1 | 1 | 0% | 2,747 | 3,320 | +21% | 0 | 0 | — |
case-15 | pass→pass | 23,597 | 27,669 | +17% | 1 | 1 | 0% | 3,198 | 4,248 | +33% | 0 | 0 | — |
case-16 | pass→pass | 19,860 | 22,895 | +15% | 1 | 1 | 0% | 2,711 | 3,736 | +38% | 0 | 0 | — |
case-17 | pass→pass | 23,396 | 29,509 | +26% | 1 | 1 | 0% | 2,990 | 4,142 | +39% | 0 | 0 | — |
case-18 | pass→pass | 27,830 | 20,856 | -25% | 1 | 1 | 0% | 2,522 | 3,544 | +41% | 0 | 0 | — |
case-19 | pass→pass | 22,383 | 27,268 | +22% | 1 | 1 | 0% | 2,902 | 4,140 | +43% | 0 | 0 | — |
case-20 | pass→pass | 29,202 | 26,926 | -8% | 1 | 1 | 0% | 3,285 | 3,941 | +20% | 0 | 0 | — |
case-21 | pass→pass | 56,156 | 21,810 | -61% | 1 | 1 | 0% | 2,661 | 3,296 | +24% | 0 | 0 | — |
case-22 | pass→pass | 20,427 | 23,531 | +15% | 1 | 1 | 0% | 2,980 | 3,625 | +22% | 0 | 0 | — |
case-23 | fail→fail | 24,222 | 21,570 | -11% | 1 | 1 | 0% | 3,250 | 4,044 | +24% | 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 +17 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.