Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →当用户需要在银行对公金融场景下，围绕客户识别进行持续监测、风险扫描、异常识别或预警提示时使用本技能。适合输出风险信号摘要、优先级判断、处置建议和升级路径。
.claude/skills/aifinlab-bank-t133-corporate-finance-customer-identification-assistant/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 127% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 289% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 80% | 0% |
| case-14 | ✗→✓ | ▲ Improved | 462% | 0% |
| case-16 | ✗→✓ | ▲ Improved | 47% | 0% |
这个 skill 面向上下游客户识别场景，重点不是直接做授信结论，而是帮助用户从交易样本、发票流水、名单和链路关系中识别关键上游、关键下游和重点排查对象。它适合用于拓客线索挖掘、链路排查和重点客户分层。
详细字段见 input-schema.md。
建议输出结构见 output-schema.md。
scripts/run_skill.py：输出上下游客户识别包shared/corporate_credit_skill_engine.py：共享分析引擎，内含客户识别专项规则| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-15 | pass→pass | 15,032 | 19,013 | +26% | 1 | 1 | 0% | 2,031 | 3,705 | +82% | 0 | 0 | — |
case-21 | pass→pass | 20,139 | 19,650 | -2% | 1 | 1 | 0% | 2,429 | 3,800 | +56% | 0 | 0 | — |
case-01 | fail→pass | 12,095 | 20,023 | +66% | 1 | 1 | 0% | 1,777 | 4,039 | +127% | 0 | 0 | — |
case-02 | pass→pass | 29,233 | 22,010 | -25% | 1 | 1 | 0% | 2,729 | 4,500 | +65% | 0 | 0 | — |
case-03 | pass→pass | 21,186 | 20,725 | -2% | 1 | 1 | 0% | 2,754 | 4,014 | +46% | 0 | 0 | — |
case-04 | pass→pass | 14,272 | 21,470 | +50% | 1 | 1 | 0% | 2,189 | 4,123 | +88% | 0 | 0 | — |
case-05 | pass→pass | 17,038 | 18,290 | +7% | 1 | 1 | 0% | 2,289 | 3,701 | +62% | 0 | 0 | — |
case-06 | fail→pass | 8,195 | 15,293 | +87% | 1 | 1 | 0% | 932 | 3,626 | +289% | 0 | 0 | — |
case-07 | pass→pass | 28,192 | 27,734 | -2% | 1 | 1 | 0% | 3,771 | 4,756 | +26% | 0 | 0 | — |
case-08 | fail→fail | 11,322 | 28,143 | +149% | 1 | 1 | 0% | 1,648 | 5,415 | +229% | 0 | 0 | — |
case-09 | fail→pass | 25,143 | 31,837 | +27% | 1 | 1 | 0% | 3,223 | 5,790 | +80% | 0 | 0 | — |
case-10 | pass→pass | 14,728 | 16,244 | +10% | 1 | 1 | 0% | 1,829 | 3,759 | +106% | 0 | 0 | — |
case-11 | pass→pass | 17,239 | 14,495 | -16% | 1 | 1 | 0% | 2,542 | 3,589 | +41% | 0 | 0 | — |
case-12 | pass→pass | 16,762 | 21,243 | +27% | 1 | 1 | 0% | 2,077 | 3,872 | +86% | 0 | 0 | — |
case-13 | fail→fail | 20,062 | 25,644 | +28% | 1 | 1 | 0% | 2,533 | 4,934 | +95% | 0 | 0 | — |
case-14 | fail→pass | 6,635 | 47,714 | +619% | 1 | 1 | 0% | 1,025 | 5,763 | +462% | 0 | 0 | — |
case-16 | fail→pass | 28,634 | 25,340 | -12% | 1 | 1 | 0% | 3,448 | 5,067 | +47% | 0 | 0 | — |
case-17 | pass→pass | 22,277 | 19,432 | -13% | 1 | 1 | 0% | 2,327 | 3,723 | +60% | 0 | 0 | — |
case-18 | pass→pass | 19,446 | 17,858 | -8% | 1 | 1 | 0% | 2,655 | 4,215 | +59% | 0 | 0 | — |
case-19 | pass→pass | 17,657 | 19,389 | +10% | 1 | 1 | 0% | 2,412 | 3,808 | +58% | 0 | 0 | — |
case-20 | pass→pass | 12,390 | 17,952 | +45% | 1 | 1 | 0% | 1,861 | 3,642 | +96% | 0 | 0 | — |
case-22 | fail→pass | 21,136 | 28,408 | +34% | 1 | 1 | 0% | 3,380 | 5,906 | +75% | 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. The headline lift of +27 percentage points is the difference between those two pass rates over the 22 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.
Other measured skills in the registry, with their headline benchmark lift.