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，对企业授信项目生成审查要点、拆解关键疑点、整理补充核验要求并形成审查岗可直接使用的结构化审阅框架时使用本技能。适合输出授信审查关注点、证据缺口、风险缓释要求和初步审查建议。
.claude/skills/aifinlab-bank-t123-corporate-finance-creditdue-diligence-review-assistant/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 114% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 140% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 100% | 0% |
| case-23 | ✓→✗ | ▼ Worse | 104% | 0% |
| case-10 | ✓→✓ | ＝ Same ✓ | 88% | 0% |
这个 skill 面向授信审查、审查支持、授信秘书或需要从审查视角复核材料的人使用。它的目标不是直接替代终审结论，而是把一笔对公授信项目拆成一套清楚、可追踪、可补证据的审查要点包。
它优先解决六类问题：
建议尽量提供：
详细字段见 input-schema.md。
默认输出至少应包含：
scripts/corporate_credit_review_points.py：生成对公授信审查要点包scripts/run_skill.py：脚本入口示例调用：
bashpython scripts/run_skill.py --input assets/example-input.json --output review_points.md --format markdown
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-10 | pass→pass | 22,556 | 27,730 | +23% | 1 | 1 | 0% | 3,054 | 5,755 | +88% | 0 | 0 | — |
case-17 | pass→pass | 24,542 | 25,372 | +3% | 1 | 1 | 0% | 3,216 | 5,896 | +83% | 0 | 0 | — |
case-01 | fail→pass | 27,676 | 44,282 | +60% | 1 | 1 | 0% | 3,625 | 7,755 | +114% | 0 | 0 | — |
case-02 | fail→pass | 15,233 | 21,713 | +43% | 1 | 1 | 0% | 2,193 | 5,273 | +140% | 0 | 0 | — |
case-03 | pass→pass | 24,362 | 30,764 | +26% | 1 | 1 | 0% | 3,637 | 6,789 | +87% | 0 | 0 | — |
case-04 | pass→pass | 11,099 | 25,709 | +132% | 1 | 1 | 0% | 1,532 | 5,341 | +249% | 0 | 0 | — |
case-05 | fail→pass | 20,244 | 26,741 | +32% | 1 | 1 | 0% | 3,143 | 6,298 | +100% | 0 | 0 | — |
case-06 | pass→pass | 28,003 | 35,597 | +27% | 1 | 1 | 0% | 3,776 | 6,887 | +82% | 0 | 0 | — |
case-07 | pass→pass | 22,415 | 55,446 | +147% | 1 | 1 | 0% | 3,151 | 6,458 | +105% | 0 | 0 | — |
case-08 | fail→fail | 26,763 | 30,427 | +14% | 1 | 1 | 0% | 3,290 | 6,850 | +108% | 0 | 0 | — |
case-09 | pass→pass | 25,049 | 29,714 | +19% | 1 | 1 | 0% | 3,011 | 6,541 | +117% | 0 | 0 | — |
case-11 | pass→pass | 19,521 | 35,223 | +80% | 1 | 1 | 0% | 2,810 | 6,310 | +125% | 0 | 0 | — |
case-12 | pass→pass | 21,649 | 31,619 | +46% | 1 | 1 | 0% | 2,577 | 6,610 | +156% | 0 | 0 | — |
case-13 | pass→pass | 19,786 | 36,247 | +83% | 1 | 1 | 0% | 2,610 | 5,304 | +103% | 0 | 0 | — |
case-14 | pass→pass | 24,454 | 33,509 | +37% | 1 | 1 | 0% | 3,343 | 6,564 | +96% | 0 | 0 | — |
case-15 | pass→pass | 21,605 | 33,570 | +55% | 1 | 1 | 0% | 3,311 | 6,373 | +92% | 0 | 0 | — |
case-16 | pass→pass | 17,119 | 25,112 | +47% | 1 | 1 | 0% | 2,368 | 5,744 | +143% | 0 | 0 | — |
case-18 | pass→pass | 20,539 | 29,390 | +43% | 1 | 1 | 0% | 2,644 | 6,845 | +159% | 0 | 0 | — |
case-19 | pass→pass | 17,456 | 26,531 | +52% | 1 | 1 | 0% | 2,524 | 5,549 | +120% | 0 | 0 | — |
case-20 | pass→pass | 27,162 | 29,782 | +10% | 1 | 1 | 0% | 2,848 | 6,638 | +133% | 0 | 0 | — |
case-21 | pass→pass | 26,019 | 33,656 | +29% | 1 | 1 | 0% | 3,564 | 7,314 | +105% | 0 | 0 | — |
case-22 | pass→pass | 21,148 | 25,745 | +22% | 1 | 1 | 0% | 2,760 | 5,494 | +99% | 0 | 0 | — |
case-23 | pass→fail | 26,680 | 31,912 | +20% | 1 | 1 | 0% | 3,166 | 6,465 | +104% | 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. 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.