Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →当需要基于制度、SOP或内控口径进行合规问答并输出可追溯的答复、引用依据和例外条件时使用本技能，适用于合规与运营支持场景。
.claude/skills/aifinlab-bank-t224-compliance-operations-policy-qa-assistant/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-03 | ✗→✓ | ▲ Improved | -2% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -20% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 25% | 0% |
| case-04 | ✓→✓ | ＝ Same ✓ | 2% | 0% |
| case-06 | ✓→✓ | ＝ Same ✓ | 24% | 0% |
本技能用于将制度条款转译为业务可执行的答复，输出清晰的规则依据、适用范围与例外条件，便于合规运营团队留痕与统一口径。
scripts/policy_qa_formatter.py| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 22,937 | 28,144 | +23% | 1 | 1 | 0% | 3,168 | 3,886 | +23% | 0 | 0 | — |
case-02 | fail→fail | 27,458 | 20,532 | -25% | 1 | 1 | 0% | 3,789 | 3,106 | -18% | 0 | 0 | — |
case-03 | fail→pass | 24,628 | 20,513 | -17% | 1 | 1 | 0% | 3,396 | 3,316 | -2% | 0 | 0 | — |
case-04 | pass→pass | 29,509 | 16,162 | -45% | 1 | 1 | 0% | 2,578 | 2,633 | +2% | 0 | 0 | — |
case-05 | fail→pass | 32,843 | 25,250 | -23% | 1 | 1 | 0% | 4,435 | 3,570 | -20% | 0 | 0 | — |
case-06 | pass→pass | 16,046 | 25,933 | +62% | 1 | 1 | 0% | 2,101 | 2,595 | +24% | 0 | 0 | — |
case-07 | fail→fail | 19,119 | 16,818 | -12% | 1 | 1 | 0% | 2,531 | 3,072 | +21% | 0 | 0 | — |
case-08 | pass→pass | 15,262 | 15,160 | -1% | 1 | 1 | 0% | 2,255 | 2,927 | +30% | 0 | 0 | — |
case-09 | fail→pass | 15,193 | 16,021 | +5% | 1 | 1 | 0% | 2,268 | 2,824 | +25% | 0 | 0 | — |
case-10 | pass→pass | 14,257 | 12,836 | -10% | 1 | 1 | 0% | 2,248 | 2,207 | -2% | 0 | 0 | — |
case-11 | pass→pass | 14,448 | 11,086 | -23% | 1 | 1 | 0% | 2,274 | 2,279 | +0% | 0 | 0 | — |
case-12 | pass→pass | 9,204 | 12,645 | +37% | 1 | 1 | 0% | 1,333 | 1,987 | +49% | 0 | 0 | — |
case-13 | pass→pass | 14,743 | 13,354 | -9% | 1 | 1 | 0% | 2,161 | 2,481 | +15% | 0 | 0 | — |
case-14 | pass→pass | 21,602 | 15,781 | -27% | 1 | 1 | 0% | 2,678 | 2,880 | +8% | 0 | 0 | — |
case-15 | pass→pass | 11,284 | 11,982 | +6% | 1 | 1 | 0% | 1,598 | 2,100 | +31% | 0 | 0 | — |
case-16 | pass→pass | 12,113 | 14,215 | +17% | 1 | 1 | 0% | 1,933 | 2,205 | +14% | 0 | 0 | — |
case-17 | pass→pass | 18,079 | 11,917 | -34% | 1 | 1 | 0% | 2,335 | 2,149 | -8% | 0 | 0 | — |
case-18 | pass→pass | 14,289 | 15,320 | +7% | 1 | 1 | 0% | 1,446 | 2,381 | +65% | 0 | 0 | — |
case-19 | pass→pass | 15,054 | 11,864 | -21% | 1 | 1 | 0% | 1,976 | 2,222 | +12% | 0 | 0 | — |
case-20 | pass→pass | 11,997 | 12,540 | +5% | 1 | 1 | 0% | 1,892 | 2,051 | +8% | 0 | 0 | — |
case-21 | pass→pass | 13,715 | 13,384 | -2% | 1 | 1 | 0% | 1,976 | 2,359 | +19% | 0 | 0 | — |
case-22 | pass→pass | 26,837 | 14,134 | -47% | 1 | 1 | 0% | 2,167 | 2,482 | +15% | 0 | 0 | — |
case-23 | pass→pass | 9,576 | 12,266 | +28% | 1 | 1 | 0% | 1,523 | 2,006 | +32% | 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 +13 percentage points is the difference between those two pass rates over the 23 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.