Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →当需要在监管报送或合规运营中定位异常、识别偏离口径的记录并输出排查路径与升级建议时使用本技能，适合异常定位和预警跟踪场景。
.claude/skills/aifinlab-bank-t222-compliance-operations-exception-localization-assistant/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 24% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -13% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 23% | 0% |
| case-12 | ✗→✓ | ▲ Improved | 14% | 0% |
| case-14 | ✗→✓ | ▲ Improved | 19% | 0% |
本技能用于对报送数据或合规运营台账中的异常进行定位、分层和排查路径设计，强调可追溯、可复核和可升级的异常处理流程。
scripts/exception_localization_report.py| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 24,058 | 25,258 | +5% | 1 | 1 | 0% | 3,495 | 4,328 | +24% | 0 | 0 | — |
case-02 | fail→fail | 28,612 | 24,717 | -14% | 1 | 1 | 0% | 4,091 | 4,048 | -1% | 0 | 0 | — |
case-03 | pass→pass | 20,622 | 18,887 | -8% | 1 | 1 | 0% | 2,430 | 2,942 | +21% | 0 | 0 | — |
case-04 | pass→pass | 13,266 | 12,328 | -7% | 1 | 1 | 0% | 1,676 | 1,987 | +19% | 0 | 0 | — |
case-05 | fail→pass | 26,547 | 16,404 | -38% | 1 | 1 | 0% | 3,305 | 2,875 | -13% | 0 | 0 | — |
case-06 | fail→fail | 28,643 | 25,169 | -12% | 1 | 1 | 0% | 3,752 | 3,734 | -0% | 0 | 0 | — |
case-07 | pass→pass | 23,884 | 23,881 | -0% | 1 | 1 | 0% | 2,885 | 3,489 | +21% | 0 | 0 | — |
case-08 | fail→fail | 23,066 | 28,171 | +22% | 1 | 1 | 0% | 2,883 | 3,705 | +29% | 0 | 0 | — |
case-09 | fail→pass | 25,708 | 24,877 | -3% | 1 | 1 | 0% | 3,086 | 3,797 | +23% | 0 | 0 | — |
case-10 | fail→fail | 25,187 | 27,929 | +11% | 1 | 1 | 0% | 3,432 | 3,905 | +14% | 0 | 0 | — |
case-11 | pass→pass | 26,398 | 26,189 | -1% | 1 | 1 | 0% | 3,330 | 4,119 | +24% | 0 | 0 | — |
case-12 | fail→pass | 26,484 | 28,184 | +6% | 1 | 1 | 0% | 3,643 | 4,156 | +14% | 0 | 0 | — |
case-13 | fail→fail | 22,695 | 25,939 | +14% | 1 | 1 | 0% | 3,181 | 3,556 | +12% | 0 | 0 | — |
case-14 | fail→pass | 24,225 | 58,431 | +141% | 1 | 1 | 0% | 3,279 | 3,909 | +19% | 0 | 0 | — |
case-15 | fail→fail | 23,804 | 20,409 | -14% | 1 | 1 | 0% | 3,138 | 3,439 | +10% | 0 | 0 | — |
case-16 | fail→pass | 29,805 | 26,281 | -12% | 1 | 1 | 0% | 3,828 | 3,952 | +3% | 0 | 0 | — |
case-17 | fail→pass | 25,011 | 24,556 | -2% | 1 | 1 | 0% | 3,106 | 3,832 | +23% | 0 | 0 | — |
case-18 | fail→pass | 22,960 | 22,401 | -2% | 1 | 1 | 0% | 2,701 | 3,679 | +36% | 0 | 0 | — |
case-19 | fail→fail | 22,716 | 19,684 | -13% | 1 | 1 | 0% | 2,747 | 3,395 | +24% | 0 | 0 | — |
case-20 | pass→pass | 26,023 | 26,672 | +2% | 1 | 1 | 0% | 3,183 | 4,059 | +28% | 0 | 0 | — |
case-21 | fail→fail | 22,947 | 21,794 | -5% | 1 | 1 | 0% | 2,972 | 3,620 | +22% | 0 | 0 | — |
case-22 | fail→fail | 25,000 | 26,583 | +6% | 1 | 1 | 0% | 3,436 | 4,289 | +25% | 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 +36 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.