Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →当用户需要在银行对公金融场景下，围绕条款核验进行完整性、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-t150-corporate-finance-task-assistant/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | 23% | 0% |
| case-16 | ✗→✓ | ▲ Improved | 109% | 0% |
| case-03 | ✓→✓ | ＝ Same ✓ | 173% | 0% |
| case-05 | ✓→✓ | ＝ Same ✓ | 30% | 0% |
| case-06 | ✓→✓ | ＝ Same ✓ | 23% | 0% |
本技能用于核验合同中与“回款确定性”强相关的条款：付款节点、账期、验收、开票、抵销/扣款、违约责任、争议解决等，并输出风险条款清单、补件与谈判要点。重点是把条款从“文本”变成“可判断/可核验/可推进”的结构化结论。
company.name + verification.time_window + verification.contractsreferences/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/corporate_ops_skill_engine.py 的 t150 场景）。
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 29,573 | 58,049 | +96% | 1 | 1 | 0% | 3,332 | 4,191 | +26% | 0 | 0 | — |
case-02 | fail→pass | 16,247 | 16,835 | +4% | 1 | 1 | 0% | 2,402 | 2,949 | +23% | 0 | 0 | — |
case-03 | pass→pass | 9,515 | 23,553 | +148% | 1 | 1 | 0% | 1,395 | 3,803 | +173% | 0 | 0 | — |
case-04 | fail→fail | 15,309 | 25,203 | +65% | 1 | 1 | 0% | 2,441 | 4,456 | +83% | 0 | 0 | — |
case-05 | pass→pass | 23,035 | 24,111 | +5% | 1 | 1 | 0% | 2,990 | 3,901 | +30% | 0 | 0 | — |
case-06 | pass→pass | 24,207 | 29,243 | +21% | 1 | 1 | 0% | 3,602 | 4,446 | +23% | 0 | 0 | — |
case-07 | pass→pass | 24,242 | 24,535 | +1% | 1 | 1 | 0% | 3,060 | 3,896 | +27% | 0 | 0 | — |
case-08 | fail→fail | 24,213 | 28,067 | +16% | 1 | 1 | 0% | 3,053 | 4,318 | +41% | 0 | 0 | — |
case-09 | pass→pass | 21,036 | 14,371 | -32% | 1 | 1 | 0% | 2,094 | 2,607 | +24% | 0 | 0 | — |
case-10 | pass→pass | 26,345 | 21,164 | -20% | 1 | 1 | 0% | 3,374 | 4,000 | +19% | 0 | 0 | — |
case-11 | pass→pass | 18,150 | 25,288 | +39% | 1 | 1 | 0% | 2,784 | 4,586 | +65% | 0 | 0 | — |
case-12 | pass→pass | 28,697 | 24,646 | -14% | 1 | 1 | 0% | 3,707 | 4,024 | +9% | 0 | 0 | — |
case-13 | pass→pass | 16,974 | 23,815 | +40% | 1 | 1 | 0% | 2,609 | 3,988 | +53% | 0 | 0 | — |
case-14 | pass→pass | 19,970 | 22,905 | +15% | 1 | 1 | 0% | 2,907 | 3,957 | +36% | 0 | 0 | — |
case-15 | pass→pass | 20,291 | 26,653 | +31% | 1 | 1 | 0% | 3,157 | 5,135 | +63% | 0 | 0 | — |
case-16 | fail→pass | 16,109 | 26,416 | +64% | 1 | 1 | 0% | 2,197 | 4,582 | +109% | 0 | 0 | — |
case-17 | pass→pass | 22,277 | 22,861 | +3% | 1 | 1 | 0% | 3,239 | 4,098 | +27% | 0 | 0 | — |
case-18 | pass→pass | 22,568 | 24,832 | +10% | 1 | 1 | 0% | 3,122 | 4,463 | +43% | 0 | 0 | — |
case-19 | pass→pass | 18,830 | 21,678 | +15% | 1 | 1 | 0% | 2,793 | 4,202 | +50% | 0 | 0 | — |
case-20 | pass→pass | 23,421 | 19,467 | -17% | 1 | 1 | 0% | 3,133 | 3,534 | +13% | 0 | 0 | — |
case-21 | pass→pass | 23,771 | 27,156 | +14% | 1 | 1 | 0% | 3,207 | 4,423 | +38% | 0 | 0 | — |
case-22 | fail→fail | 13,204 | 18,957 | +44% | 1 | 1 | 0% | 1,997 | 3,705 | +86% | 0 | 0 | — |
case-23 | fail→fail | 23,650 | 21,364 | -10% | 1 | 1 | 0% | 2,940 | 3,861 | +31% | 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 +9 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.