Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use when you need a bank corporate credit issue-list assistant (问题清单/补件清单/催办台账). Trigger this skill for把审查问题拆成“严重程度、证据要求、责任到人、截止时间、升级路径”的可执行清单，并可用脚本从结构化输入生成问题台账与催办要点。
.claude/skills/aifinlab-bank-t143-corporate-finance-task-assistant/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | 85% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 46% | 0% |
| case-12 | ✗→✓ | ▲ Improved | 53% | 0% |
| case-13 | ✗→✓ | ▲ Improved | 39% | 0% |
| case-09 | ✓→✗ | ▼ Worse | 76% | 0% |
本技能把“审查提问/补件要求/风险关注点”整理成可推进的台账：
最低必需（详见 references/input-schema.md）：
company.nameissue_context.stage（准入/续作/贷后/重组等）issues[]（每条至少包含 issue 与 severity）强烈建议：
issues[].required_materials：最小证据清单issues[].owner 与 issues[].due_date：便于催办闭环脚本输出字段见 references/output-schema.md。
bashpython scripts/run_skill.py --input assets/example-input.json --format markdown python scripts/run_skill.py --input assets/example-input.json --format json
scripts/run_skill.py：命令行入口scripts/credit_issue_list_builder.py：t143 场景封装（调用共享引擎）| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 25,102 | 33,463 | +33% | 1 | 1 | 0% | 3,367 | 4,723 | +40% | 0 | 0 | — |
case-10 | pass→pass | 16,422 | 32,066 | +95% | 1 | 1 | 0% | 2,368 | 4,938 | +109% | 0 | 0 | — |
case-02 | fail→pass | 16,625 | 22,446 | +35% | 1 | 1 | 0% | 2,432 | 4,488 | +85% | 0 | 0 | — |
case-03 | pass→pass | 4,956 | 16,366 | +230% | 1 | 1 | 0% | 588 | 3,211 | +446% | 0 | 0 | — |
case-04 | pass→pass | 15,267 | 53,877 | +253% | 1 | 1 | 0% | 2,185 | 4,352 | +99% | 0 | 0 | — |
case-05 | fail→fail | 18,065 | 19,186 | +6% | 1 | 1 | 0% | 2,750 | 4,164 | +51% | 0 | 0 | — |
case-06 | fail→pass | 17,236 | 21,053 | +22% | 1 | 1 | 0% | 2,673 | 3,912 | +46% | 0 | 0 | — |
case-07 | pass→pass | 16,029 | 25,322 | +58% | 1 | 1 | 0% | 2,477 | 4,675 | +89% | 0 | 0 | — |
case-08 | pass→pass | 49,967 | 27,945 | -44% | 1 | 1 | 0% | 3,618 | 4,649 | +28% | 0 | 0 | — |
case-09 | pass→fail | 21,630 | 27,524 | +27% | 1 | 1 | 0% | 2,632 | 4,637 | +76% | 0 | 0 | — |
case-11 | pass→pass | 28,570 | 25,659 | -10% | 1 | 1 | 0% | 3,821 | 4,757 | +24% | 0 | 0 | — |
case-12 | fail→pass | 24,949 | 25,659 | +3% | 1 | 1 | 0% | 3,284 | 5,025 | +53% | 0 | 0 | — |
case-13 | fail→pass | 25,819 | 28,636 | +11% | 1 | 1 | 0% | 3,429 | 4,778 | +39% | 0 | 0 | — |
case-14 | fail→fail | 18,431 | 22,815 | +24% | 1 | 1 | 0% | 2,941 | 4,521 | +54% | 0 | 0 | — |
case-15 | pass→pass | 17,454 | 16,645 | -5% | 1 | 1 | 0% | 2,126 | 3,687 | +73% | 0 | 0 | — |
case-16 | pass→pass | 13,897 | 22,459 | +62% | 1 | 1 | 0% | 1,810 | 4,175 | +131% | 0 | 0 | — |
case-17 | pass→pass | 20,660 | 27,182 | +32% | 1 | 1 | 0% | 3,157 | 5,230 | +66% | 0 | 0 | — |
case-18 | pass→pass | 20,415 | 22,308 | +9% | 1 | 1 | 0% | 3,087 | 4,602 | +49% | 0 | 0 | — |
case-19 | pass→pass | 23,230 | 42,069 | +81% | 1 | 1 | 0% | 3,208 | 5,055 | +58% | 0 | 0 | — |
case-20 | pass→pass | 13,526 | 21,569 | +59% | 1 | 1 | 0% | 1,818 | 3,806 | +109% | 0 | 0 | — |
case-21 | fail→fail | 15,778 | 24,698 | +57% | 1 | 1 | 0% | 2,381 | 4,737 | +99% | 0 | 0 | — |
case-22 | pass→pass | 20,397 | 29,335 | +44% | 1 | 1 | 0% | 2,894 | 4,982 | +72% | 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 +14 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.