Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use this skill when the user asks for help with corporate credit due diligence, enterprise credit investigation, pre-loan review, borrower analysis, document collection, management interview preparation, risk identification, or drafting due diligence reports in banking corporate finance scenarios. Trigger this skill for requests involving: - 对公授信尽调 - 企业客户授信调查 - 贷前调查与审查 - 授信资料清单整理 - 企业经营与财务风险识别 - 尽调报告、访谈提纲、风险提示、授信分析摘要生成 Do not use this skill for retail banking, consumer lending, personal mortga
.claude/skills/aifinlab-bank-corporate-credit-dd/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | 41% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 5% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 5% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 83% | 0% |
| case-17 | ✗→✓ | ▲ Improved | 53% | 0% |
This skill helps the assistant support banking corporate credit due diligence tasks. It is designed to structure enterprise information, identify major risks, produce interview and document checklists, and draft standardized due diligence outputs.
This skill supports:
This skill does not replace:
Use this skill when the user wants to:
Do not use this skill when:
Always distinguish among:
Never present assumptions as verified facts. When evidence is insufficient, explicitly mark items as:
Determine whether the user needs one of these:
Organize available information into:
Identify missing critical materials such as:
Review risks across multiple dimensions:
Depending on user request, generate:
If information is incomplete, provide conditional views instead of final approval recommendations.
Outputs should usually include these sections when relevant:
For each risk item, include:
Never help the user:
Read references/due-diligence-checklist.md for material checklist. Read references/risk-red-flags.md for common banking red flags. Read references/report-template.md for standard report structure. Read references/interview-outline-template.md for management interview questions. Read references/output-schema.md for structured output format.
python# 调用 skill result = run_skill({ "param1": "value1", "param2": "value2" })
bashpython scripts/run_skill.py --input data.json
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 21,215 | 35,272 | +66% | 1 | 1 | 0% | 3,285 | 5,537 | +69% | 0 | 0 | — |
case-02 | fail→pass | 23,827 | 25,219 | +6% | 1 | 1 | 0% | 3,397 | 4,783 | +41% | 0 | 0 | — |
case-03 | fail→pass | 30,787 | 17,244 | -44% | 1 | 1 | 0% | 3,674 | 3,857 | +5% | 0 | 0 | — |
case-04 | fail→pass | 30,544 | 19,030 | -38% | 1 | 1 | 0% | 3,980 | 4,196 | +5% | 0 | 0 | — |
case-05 | pass→pass | 11,954 | 8,545 | -29% | 1 | 1 | 0% | 1,625 | 2,559 | +57% | 0 | 0 | — |
case-21 | pass→pass | 19,698 | 21,011 | +7% | 1 | 1 | 0% | 2,715 | 4,365 | +61% | 0 | 0 | — |
case-06 | fail→fail | 14,446 | 16,211 | +12% | 1 | 1 | 0% | 2,105 | 3,700 | +76% | 0 | 0 | — |
case-07 | pass→pass | 28,747 | 22,916 | -20% | 1 | 1 | 0% | 3,358 | 4,503 | +34% | 0 | 0 | — |
case-08 | fail→pass | 22,141 | 29,183 | +32% | 1 | 1 | 0% | 3,086 | 5,651 | +83% | 0 | 0 | — |
case-09 | pass→pass | 42,382 | 30,377 | -28% | 1 | 1 | 0% | 4,830 | 5,292 | +10% | 0 | 0 | — |
case-10 | pass→pass | 18,174 | 20,413 | +12% | 1 | 1 | 0% | 2,552 | 4,250 | +67% | 0 | 0 | — |
case-11 | pass→pass | 27,453 | 27,785 | +1% | 1 | 1 | 0% | 3,790 | 4,743 | +25% | 0 | 0 | — |
case-12 | pass→pass | 21,062 | 22,697 | +8% | 1 | 1 | 0% | 2,857 | 4,645 | +63% | 0 | 0 | — |
case-13 | pass→pass | 30,540 | 22,013 | -28% | 1 | 1 | 0% | 2,593 | 4,022 | +55% | 0 | 0 | — |
case-14 | pass→pass | 15,542 | 19,952 | +28% | 1 | 1 | 0% | 2,297 | 4,472 | +95% | 0 | 0 | — |
case-15 | fail→fail | 27,998 | 29,974 | +7% | 1 | 1 | 0% | 3,699 | 5,453 | +47% | 0 | 0 | — |
case-16 | pass→pass | 27,518 | 27,250 | -1% | 1 | 1 | 0% | 3,427 | 4,764 | +39% | 0 | 0 | — |
case-17 | fail→pass | 19,605 | 26,979 | +38% | 1 | 1 | 0% | 3,137 | 4,805 | +53% | 0 | 0 | — |
case-18 | pass→pass | 20,854 | 24,074 | +15% | 1 | 1 | 0% | 2,723 | 4,378 | +61% | 0 | 0 | — |
case-19 | pass→pass | 22,512 | 28,899 | +28% | 1 | 1 | 0% | 3,005 | 4,983 | +66% | 0 | 0 | — |
case-20 | pass→pass | 25,613 | 12,837 | -50% | 1 | 1 | 0% | 3,328 | 3,340 | +0% | 0 | 0 | — |
case-22 | pass→pass | 22,120 | 33,490 | +51% | 1 | 1 | 0% | 3,349 | 6,038 | +80% | 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.
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.