Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →当用户需要在银行零售金融场景下，对信用卡客户的流失风险进行持续监测、信号识别、分层预警与行动建议时使用本技能。适合输出预警名单、风险级别、触达优先级与挽留动作建议。
.claude/skills/aifinlab-bank-t166-retail-finance-card-churn-warning-assistant/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-13 | ✗→✓ | ▲ Improved | -19% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 62% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -16% | 0% |
| case-16 | ✗→✓ | ▲ Improved | 31% | 0% |
| case-18 | ✗→✓ | ▲ Improved | 29% | 0% |
本技能用于把信用卡客户的“流失信号”转化成可执行的预警名单与挽留动作包。强调客群分层、信号解释、行动优先级和合规触达边界，不替代审批或收益承诺。
batch_id、time_window、customersreferences/input-schema.mdbashpython scripts/run_skill.py --input assets/example-input.json --format markdown python scripts/run_skill.py --input assets/example-input.json --format json
脚本入口：scripts/run_skill.py（调用 shared/retail_service_ops_skill_engine.py 的 t166 场景）。
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-13 | fail→pass | 8,610 | 5,044 | -41% | 1 | 1 | 0% | 1,304 | 1,056 | -19% | 0 | 0 | — |
case-01 | fail→fail | 59,490 | 30,449 | -49% | 1 | 1 | 0% | 4,186 | 4,527 | +8% | 0 | 0 | — |
case-02 | fail→fail | 25,023 | 36,321 | +45% | 1 | 1 | 0% | 3,684 | 4,087 | +11% | 0 | 0 | — |
case-03 | fail→pass | 12,881 | 21,482 | +67% | 1 | 1 | 0% | 2,005 | 3,243 | +62% | 0 | 0 | — |
case-04 | pass→pass | 15,212 | 21,008 | +38% | 1 | 1 | 0% | 2,264 | 3,708 | +64% | 0 | 0 | — |
case-05 | fail→pass | 28,025 | 25,266 | -10% | 1 | 1 | 0% | 3,985 | 3,337 | -16% | 0 | 0 | — |
case-06 | pass→pass | 16,859 | 19,185 | +14% | 1 | 1 | 0% | 2,642 | 3,141 | +19% | 0 | 0 | — |
case-07 | pass→pass | 15,612 | 14,937 | -4% | 1 | 1 | 0% | 2,376 | 2,964 | +25% | 0 | 0 | — |
case-08 | pass→pass | 18,167 | 23,651 | +30% | 1 | 1 | 0% | 2,785 | 3,734 | +34% | 0 | 0 | — |
case-09 | pass→pass | 17,337 | 20,216 | +17% | 1 | 1 | 0% | 2,514 | 3,286 | +31% | 0 | 0 | — |
case-10 | pass→pass | 22,388 | 18,406 | -18% | 1 | 1 | 0% | 2,894 | 3,125 | +8% | 0 | 0 | — |
case-11 | pass→pass | 10,889 | 14,508 | +33% | 1 | 1 | 0% | 1,618 | 2,642 | +63% | 0 | 0 | — |
case-12 | pass→pass | 21,582 | 26,320 | +22% | 1 | 1 | 0% | 3,110 | 4,166 | +34% | 0 | 0 | — |
case-14 | pass→pass | 21,192 | 28,083 | +33% | 1 | 1 | 0% | 2,779 | 4,197 | +51% | 0 | 0 | — |
case-15 | pass→pass | 20,989 | 25,166 | +20% | 1 | 1 | 0% | 2,519 | 4,138 | +64% | 0 | 0 | — |
case-16 | fail→pass | 19,405 | 23,471 | +21% | 1 | 1 | 0% | 2,895 | 3,781 | +31% | 0 | 0 | — |
case-17 | pass→pass | 16,031 | 18,750 | +17% | 1 | 1 | 0% | 2,151 | 3,228 | +50% | 0 | 0 | — |
case-18 | fail→pass | 21,864 | 21,382 | -2% | 1 | 1 | 0% | 2,870 | 3,716 | +29% | 0 | 0 | — |
case-19 | pass→pass | 21,901 | 27,383 | +25% | 1 | 1 | 0% | 2,527 | 4,194 | +66% | 0 | 0 | — |
case-20 | fail→pass | 21,460 | 4,836 | -77% | 1 | 1 | 0% | 1,701 | 1,226 | -28% | 0 | 0 | — |
case-21 | fail→pass | 20,985 | 21,467 | +2% | 1 | 1 | 0% | 3,078 | 3,535 | +15% | 0 | 0 | — |
case-22 | fail→pass | 24,260 | 21,661 | -11% | 1 | 1 | 0% | 3,182 | 3,452 | +8% | 0 | 0 | — |
case-23 | pass→pass | 17,265 | 15,878 | -8% | 1 | 1 | 0% | 2,254 | 2,990 | +33% | 0 | 0 | — |
case-24 | pass→pass | 17,943 | 20,972 | +17% | 1 | 1 | 0% | 2,318 | 3,367 | +45% | 0 | 0 | — |
case-25 | fail→pass | 24,197 | 18,520 | -23% | 1 | 1 | 0% | 3,215 | 3,228 | +0% | 0 | 0 | — |
case-26 | pass→pass | 18,368 | 14,451 | -21% | 1 | 1 | 0% | 2,393 | 3,048 | +27% | 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. 26 cases were attempted. The headline lift of +35 percentage points is the difference between those two pass rates over the 26 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.