Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →当需要对监管报送数据或材料做口径校验、完整性/一致性检查并输出异常清单与补件要求时使用本技能，适合合规与运营报送团队日常核验场景。
.claude/skills/aifinlab-bank-t221-compliance-operations-reporting-validation-assistant/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-07 | ✗→✓ | ▲ Improved | 29% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 79% | 0% |
| case-19 | ✗→✓ | ▲ Improved | 23% | 0% |
| case-21 | ✗→✓ | ▲ Improved | -1% | 0% |
| case-03 | ✓→✓ | ＝ Same ✓ | 4% | 0% |
本技能用于把监管报送规则、字段口径与源数据进行对照核验，输出可追溯的校验结论、异常定位和补件建议，适合报送前检查和内部质检。
scripts/reporting_validation_checker.py| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-02 | fail→fail | 16,697 | 18,975 | +14% | 1 | 1 | 0% | 2,591 | 2,974 | +15% | 0 | 0 | — |
case-03 | pass→pass | 22,629 | 15,546 | -31% | 1 | 1 | 0% | 2,704 | 2,821 | +4% | 0 | 0 | — |
case-04 | pass→pass | 18,248 | 10,125 | -45% | 1 | 1 | 0% | 2,351 | 1,901 | -19% | 0 | 0 | — |
case-05 | pass→pass | 16,472 | 15,325 | -7% | 1 | 1 | 0% | 2,112 | 2,666 | +26% | 0 | 0 | — |
case-06 | pass→pass | 19,851 | 20,223 | +2% | 1 | 1 | 0% | 3,121 | 3,120 | -0% | 0 | 0 | — |
case-07 | fail→pass | 23,668 | 35,365 | +49% | 1 | 1 | 0% | 3,438 | 4,440 | +29% | 0 | 0 | — |
case-08 | fail→pass | 13,253 | 21,296 | +61% | 1 | 1 | 0% | 1,748 | 3,130 | +79% | 0 | 0 | — |
case-09 | pass→pass | 19,177 | 18,314 | -5% | 1 | 1 | 0% | 2,392 | 3,191 | +33% | 0 | 0 | — |
case-10 | fail→fail | 17,059 | 15,091 | -12% | 1 | 1 | 0% | 3,253 | 3,294 | +1% | 0 | 0 | — |
case-11 | pass→pass | 15,230 | 18,592 | +22% | 1 | 1 | 0% | 2,457 | 3,180 | +29% | 0 | 0 | — |
case-01 | fail→fail | 32,583 | 31,010 | -5% | 1 | 1 | 0% | 4,502 | 4,795 | +7% | 0 | 0 | — |
case-12 | pass→pass | 20,288 | 20,167 | -1% | 1 | 1 | 0% | 3,021 | 3,375 | +12% | 0 | 0 | — |
case-13 | pass→pass | 26,785 | 21,885 | -18% | 1 | 1 | 0% | 2,757 | 3,296 | +20% | 0 | 0 | — |
case-14 | pass→pass | 13,353 | 13,387 | +0% | 1 | 1 | 0% | 1,862 | 2,571 | +38% | 0 | 0 | — |
case-15 | pass→pass | 24,653 | 20,902 | -15% | 1 | 1 | 0% | 3,078 | 3,103 | +1% | 0 | 0 | — |
case-16 | pass→pass | 13,911 | 18,675 | +34% | 1 | 1 | 0% | 2,016 | 2,977 | +48% | 0 | 0 | — |
case-17 | pass→pass | 11,453 | 14,596 | +27% | 1 | 1 | 0% | 1,715 | 2,837 | +65% | 0 | 0 | — |
case-18 | pass→pass | 17,058 | 18,032 | +6% | 1 | 1 | 0% | 2,295 | 3,167 | +38% | 0 | 0 | — |
case-19 | fail→pass | 18,812 | 17,660 | -6% | 1 | 1 | 0% | 2,415 | 2,967 | +23% | 0 | 0 | — |
case-20 | pass→pass | 11,594 | 13,282 | +15% | 1 | 1 | 0% | 1,603 | 2,252 | +40% | 0 | 0 | — |
case-21 | fail→pass | 25,782 | 20,884 | -19% | 1 | 1 | 0% | 3,709 | 3,660 | -1% | 0 | 0 | — |
case-22 | pass→pass | 32,929 | 22,286 | -32% | 1 | 1 | 0% | 3,032 | 3,516 | +16% | 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 +18 percentage points is the difference between those two pass rates over the 22 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.