Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →汇总客户分层、流失预警、产品渗透、沉睡资金、提前还款等多维分析结果，生成可执行的客户经营机会清单与优先级排序，对接客户经理日常经营。
.claude/skills/aifinlab-customer-opportunity-list-generation/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-03 | ✗→✓ | ▲ Improved | -28% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -35% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -26% | 0% |
| case-12 | ✗→✓ | ▲ Improved | 29% | 0% |
| case-17 | ✗→✓ | ▲ Improved | 10% | 0% |
银行各类客户分析产出分散在不同系统和报告中，客户经理难以快速获取汇总后的可执行经营动作。需要将客户分层、流失预警、产品渗透、沉睡资金激活、提前还款挽留等多维分析结果整合为一份优先级排序的机会清单，直接对接客户经理日常经营工作。
| 输入项 | 说明 | 是否必填 | |--------|------|----------| | 客户分层结果 | 来自 customer-segmentation 的分层数据 | 选填 | | 流失预警结果 | 来自 customer-churn-alert 的预警清单 | 选填 | | 产品渗透分析 | 来自 customer-product-penetration 的交叉销售机会 | 选填 | | 沉睡资金分析 | 来自 dormant-account-analysis 的激活机会 | 选填 | | 提前还款分析 | 来自 early-repayment-churn-analysis 的挽留对象 | 选填 | | 客户经理归属 | 客户与客户经理的对应关系 | 选填 | | 优先级规则 | 自定义优先级权重（默认使用内置规则） | 选填 |
备注：至少需要一项上游分析结果作为输入。
无（终端输出技能，直接对接客户经理日常经营工作流）
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-13 | pass→fail | 16,720 | 20,625 | +23% | 1 | 1 | 0% | 2,426 | 4,161 | +72% | 0 | 0 | — |
case-14 | pass→pass | 12,592 | 12,445 | -1% | 1 | 1 | 0% | 1,755 | 2,740 | +56% | 0 | 0 | — |
case-01 | fail→fail | 23,073 | 47,017 | +104% | 1 | 1 | 0% | 3,750 | 5,445 | +45% | 0 | 0 | — |
case-02 | fail→fail | 26,262 | 24,828 | -5% | 1 | 1 | 0% | 4,185 | 5,171 | +24% | 0 | 0 | — |
case-03 | fail→pass | 24,168 | 9,961 | -59% | 1 | 1 | 0% | 3,431 | 2,473 | -28% | 0 | 0 | — |
case-04 | pass→pass | 20,679 | 18,670 | -10% | 1 | 1 | 0% | 3,156 | 3,833 | +21% | 0 | 0 | — |
case-05 | pass→pass | 19,596 | 10,329 | -47% | 1 | 1 | 0% | 2,804 | 2,464 | -12% | 0 | 0 | — |
case-06 | pass→pass | 15,645 | 14,088 | -10% | 1 | 1 | 0% | 2,586 | 3,682 | +42% | 0 | 0 | — |
case-07 | fail→pass | 15,052 | 3,087 | -79% | 1 | 1 | 0% | 2,312 | 1,492 | -35% | 0 | 0 | — |
case-08 | fail→pass | 13,028 | 2,549 | -80% | 1 | 1 | 0% | 1,941 | 1,431 | -26% | 0 | 0 | — |
case-09 | pass→pass | 15,427 | 13,093 | -15% | 1 | 1 | 0% | 2,139 | 2,963 | +39% | 0 | 0 | — |
case-10 | pass→pass | 11,070 | 10,977 | -1% | 1 | 1 | 0% | 1,904 | 2,904 | +53% | 0 | 0 | — |
case-11 | pass→pass | 21,405 | 18,601 | -13% | 1 | 1 | 0% | 3,251 | 3,935 | +21% | 0 | 0 | — |
case-12 | fail→pass | 15,509 | 11,727 | -24% | 1 | 1 | 0% | 2,324 | 2,989 | +29% | 0 | 0 | — |
case-15 | pass→pass | 17,002 | 3,296 | -81% | 1 | 1 | 0% | 2,558 | 1,523 | -40% | 0 | 0 | — |
case-16 | fail→fail | 21,036 | 18,754 | -11% | 1 | 1 | 0% | 2,945 | 3,442 | +17% | 0 | 0 | — |
case-17 | fail→pass | 11,831 | 3,368 | -72% | 1 | 1 | 0% | 1,436 | 1,575 | +10% | 0 | 0 | — |
case-18 | pass→pass | 14,159 | 9,278 | -34% | 1 | 1 | 0% | 2,088 | 2,588 | +24% | 0 | 0 | — |
case-19 | pass→pass | 17,409 | 15,825 | -9% | 1 | 1 | 0% | 2,457 | 3,200 | +30% | 0 | 0 | — |
case-20 | pass→pass | 10,809 | 6,513 | -40% | 1 | 1 | 0% | 1,587 | 1,971 | +24% | 0 | 0 | — |
case-21 | fail→pass | 15,594 | 10,323 | -34% | 1 | 1 | 0% | 2,197 | 2,628 | +20% | 0 | 0 | — |
case-22 | pass→pass | 14,125 | 13,978 | -1% | 1 | 1 | 0% | 2,013 | 2,903 | +44% | 0 | 0 | — |
case-23 | fail→pass | 10,589 | 4,500 | -58% | 1 | 1 | 0% | 1,804 | 1,792 | -1% | 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 +26 percentage points is the difference between those two pass rates over the 23 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.