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，以客户经理视角推进企业授信尽调、整理资料清单、设计访谈提纲、识别现场核查重点、形成风险假设和输出尽调纪要时使用本技能。适合输出能直接支持客户拜访和授信推进的尽调包。
.claude/skills/aifinlab-bank-t122-corporate-finance-creditdue-diligence-rm-assistant/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-03 | ✗→✓ | ▲ Improved | 52% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 93% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 141% | 0% |
| case-18 | ✗→✓ | ▲ Improved | 136% | 0% |
| case-19 | ✗→✓ | ▲ Improved | 70% | 0% |
这个 skill 面向一线客户经理使用。它不是审查岗的终审工具，也不是审批会的拍板工具，而是帮助客户经理把授信尽调从“到处找材料、临时想问题”变成一套可执行的推进动作。
它优先解决五类问题：
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_dd_rm.py：生成客户经理版尽调动作包scripts/run_skill.py：脚本入口示例调用：
bashpython scripts/run_skill.py --input assets/example-input.json --output dd_packet.md --format markdown
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 30,514 | 52,469 | +72% | 1 | 1 | 0% | 4,402 | 7,190 | +63% | 0 | 0 | — |
case-02 | fail→fail | 28,810 | 41,591 | +44% | 1 | 1 | 0% | 3,932 | 8,037 | +104% | 0 | 0 | — |
case-03 | fail→pass | 29,936 | 31,544 | +5% | 1 | 1 | 0% | 4,400 | 6,703 | +52% | 0 | 0 | — |
case-04 | fail→pass | 18,402 | 19,925 | +8% | 1 | 1 | 0% | 2,595 | 5,008 | +93% | 0 | 0 | — |
case-05 | pass→pass | 19,170 | 21,248 | +11% | 1 | 1 | 0% | 2,411 | 5,158 | +114% | 0 | 0 | — |
case-06 | fail→fail | 23,260 | 44,855 | +93% | 1 | 1 | 0% | 3,641 | 7,894 | +117% | 0 | 0 | — |
case-07 | fail→pass | 19,117 | 33,065 | +73% | 1 | 1 | 0% | 2,684 | 6,461 | +141% | 0 | 0 | — |
case-08 | pass→pass | 23,624 | 27,597 | +17% | 1 | 1 | 0% | 3,352 | 5,464 | +63% | 0 | 0 | — |
case-09 | pass→pass | 19,883 | 26,465 | +33% | 1 | 1 | 0% | 2,716 | 5,307 | +95% | 0 | 0 | — |
case-10 | fail→fail | 36,994 | 44,465 | +20% | 1 | 1 | 0% | 3,749 | 7,390 | +97% | 0 | 0 | — |
case-11 | fail→fail | 31,368 | 64,751 | +106% | 1 | 1 | 0% | 3,997 | 7,055 | +77% | 0 | 0 | — |
case-12 | pass→pass | 21,907 | 36,514 | +67% | 1 | 1 | 0% | 3,217 | 6,408 | +99% | 0 | 0 | — |
case-13 | pass→pass | 35,231 | 28,151 | -20% | 1 | 1 | 0% | 2,998 | 5,907 | +97% | 0 | 0 | — |
case-14 | fail→fail | 31,442 | 33,640 | +7% | 1 | 1 | 0% | 3,597 | 6,209 | +73% | 0 | 0 | — |
case-15 | pass→pass | 25,923 | 35,717 | +38% | 1 | 1 | 0% | 3,264 | 6,741 | +107% | 0 | 0 | — |
case-16 | pass→pass | 22,729 | 28,856 | +27% | 1 | 1 | 0% | 3,077 | 6,197 | +101% | 0 | 0 | — |
case-17 | pass→pass | 22,145 | 26,250 | +19% | 1 | 1 | 0% | 2,747 | 5,182 | +89% | 0 | 0 | — |
case-18 | fail→pass | 22,421 | 31,521 | +41% | 1 | 1 | 0% | 2,752 | 6,507 | +136% | 0 | 0 | — |
case-19 | fail→pass | 27,806 | 28,684 | +3% | 1 | 1 | 0% | 3,381 | 5,742 | +70% | 0 | 0 | — |
case-20 | fail→pass | 22,239 | 31,946 | +44% | 1 | 1 | 0% | 3,077 | 6,031 | +96% | 0 | 0 | — |
case-21 | pass→fail | 30,462 | 55,933 | +84% | 1 | 1 | 0% | 3,706 | 7,728 | +109% | 0 | 0 | — |
case-22 | fail→fail | 24,641 | 52,145 | +112% | 1 | 1 | 0% | 3,470 | 8,675 | +150% | 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 +23 percentage points is the difference between those two pass rates over the 22 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.