Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →当用户需要在银行交易银行与普惠场景下，围绕流水核验进行完整性、一致性、真实性或规则符合性检查时使用本技能。适合输出核验结论、异常项清单、补件要求和升级复核建议。
.claude/skills/aifinlab-bank-t242-transaction-banking-inclusive-finance-task-assistant/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | 97% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -15% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 39% | 0% |
| case-23 | ✗→✓ | ▲ Improved | 20% | 0% |
| case-08 | ✓→✓ | ＝ Same ✓ | 81% | 0% |
本技能面向交易银行与普惠金融场景，既要兼顾准入效率和材料真实性，也要把结算、现金管理、供应链和贸易融资方案写到能落地的层面。 当前这支 skill 更偏向服务 交易银行和普惠团队，输出时要特别注意 兼顾准入效率、方案落地和风险边界。
本技能提供账户流水核验与异常清单脚本，支持重复输出核验结论与补件建议。
scripts/flow_validation_report.py用途：对账户流水进行完整性、一致性、异常频率与大额异动识别。
输入字段（JSON）：
accounts: 账户清单cash_flows: 流水明细rules: 阈值规则（大额、频率、对手）period: 样本期间输出内容：
命令行示例：
python scripts/flow_validation_report.py --input assets/flows.json --output outputs/flow_validation.md| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-08 | pass→pass | 16,472 | 22,289 | +35% | 1 | 1 | 0% | 2,303 | 4,163 | +81% | 0 | 0 | — |
case-01 | fail→fail | 53,676 | 64,640 | +20% | 1 | 1 | 0% | 3,350 | 5,523 | +65% | 0 | 0 | — |
case-02 | fail→pass | 14,036 | 23,017 | +64% | 1 | 1 | 0% | 2,032 | 4,007 | +97% | 0 | 0 | — |
case-03 | pass→pass | 6,094 | 45,228 | +642% | 1 | 1 | 0% | 811 | 3,122 | +285% | 0 | 0 | — |
case-04 | pass→pass | 48,833 | 19,673 | -60% | 1 | 1 | 0% | 2,321 | 3,740 | +61% | 0 | 0 | — |
case-05 | fail→pass | 13,382 | 5,971 | -55% | 1 | 1 | 0% | 2,254 | 1,921 | -15% | 0 | 0 | — |
case-06 | fail→pass | 48,065 | 47,930 | -0% | 1 | 1 | 0% | 2,594 | 3,607 | +39% | 0 | 0 | — |
case-07 | fail→fail | 19,473 | 54,790 | +181% | 1 | 1 | 0% | 2,750 | 4,128 | +50% | 0 | 0 | — |
case-09 | pass→pass | 14,294 | 19,335 | +35% | 1 | 1 | 0% | 2,256 | 3,938 | +75% | 0 | 0 | — |
case-10 | pass→pass | 6,491 | 2,663 | -59% | 1 | 1 | 0% | 1,244 | 1,492 | +20% | 0 | 0 | — |
case-11 | pass→pass | 20,271 | 27,280 | +35% | 1 | 1 | 0% | 2,750 | 4,826 | +75% | 0 | 0 | — |
case-12 | pass→pass | 20,238 | 21,692 | +7% | 1 | 1 | 0% | 2,827 | 4,057 | +44% | 0 | 0 | — |
case-13 | pass→pass | 24,414 | 22,776 | -7% | 1 | 1 | 0% | 3,178 | 4,266 | +34% | 0 | 0 | — |
case-14 | pass→pass | 14,748 | 11,190 | -24% | 1 | 1 | 0% | 2,200 | 2,674 | +22% | 0 | 0 | — |
case-15 | pass→pass | 18,842 | 23,778 | +26% | 1 | 1 | 0% | 2,695 | 4,380 | +63% | 0 | 0 | — |
case-16 | pass→pass | 21,399 | 26,858 | +26% | 1 | 1 | 0% | 2,960 | 4,950 | +67% | 0 | 0 | — |
case-17 | pass→pass | 22,691 | 26,135 | +15% | 1 | 1 | 0% | 3,236 | 4,856 | +50% | 0 | 0 | — |
case-18 | pass→pass | 19,117 | 24,180 | +26% | 1 | 1 | 0% | 2,658 | 4,368 | +64% | 0 | 0 | — |
case-19 | pass→pass | 21,243 | 19,231 | -9% | 1 | 1 | 0% | 2,907 | 3,990 | +37% | 0 | 0 | — |
case-20 | pass→pass | 26,345 | 25,418 | -4% | 1 | 1 | 0% | 3,472 | 4,589 | +32% | 0 | 0 | — |
case-21 | fail→fail | 15,310 | 21,174 | +38% | 1 | 1 | 0% | 2,166 | 3,919 | +81% | 0 | 0 | — |
case-22 | pass→pass | 15,789 | 19,909 | +26% | 1 | 1 | 0% | 2,245 | 3,716 | +66% | 0 | 0 | — |
case-23 | fail→pass | 15,887 | 13,303 | -16% | 1 | 1 | 0% | 2,407 | 2,882 | +20% | 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. 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.