Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use when you need a bank corporate client operating-volatility monitoring assistant (经营监测/指标波动/预警解释). Trigger this skill for requests to把 KPI 序列与告警整理成“异常信号、可能驱动、核验材料、处置建议、对客解释稿”，并可用脚本从结构化输入生成周报/快报。
.claude/skills/aifinlab-bank-t141-corporate-finance-task-assistant/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-03 | ✗→✓ | ▲ Improved | 100% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 68% | 0% |
| case-11 | ✗→✓ | ▲ Improved | 58% | 0% |
| case-17 | ✗→✓ | ▲ Improved | 65% | 0% |
| case-21 | ✗→✓ | ▲ Improved | 41% | 0% |
本技能用于把“经营指标波动”转成可核验、可沟通、可处置的监测输出。
核心目标：
典型输入：
典型输出：
最低必需（详见 references/input-schema.md）：
company.namemonitoring.time_windowmonitoring.kpis[]：至少 2 期（推荐 3-6 期），包含 period 与关键指标（收入/利润/现金流）强烈建议：
monitoring.kpi_definitions：口径说明monitoring.baseline：历史基线/同业对标（用于判断行业共振）monitoring.alerts[]：系统告警或人工发现脚本输出字段见 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/operating_volatility_monitor.py：t141 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 | 26,360 | 61,461 | +133% | 1 | 1 | 0% | 4,468 | 7,568 | +69% | 0 | 0 | — |
case-02 | pass→pass | 25,625 | 28,734 | +12% | 1 | 1 | 0% | 3,633 | 5,877 | +62% | 0 | 0 | — |
case-03 | fail→pass | 22,212 | 29,909 | +35% | 1 | 1 | 0% | 3,027 | 6,046 | +100% | 0 | 0 | — |
case-04 | fail→pass | 24,689 | 30,945 | +25% | 1 | 1 | 0% | 3,266 | 5,489 | +68% | 0 | 0 | — |
case-05 | pass→pass | 29,902 | 33,394 | +12% | 1 | 1 | 0% | 3,563 | 6,569 | +84% | 0 | 0 | — |
case-06 | pass→pass | 29,768 | 31,732 | +7% | 1 | 1 | 0% | 3,975 | 5,566 | +40% | 0 | 0 | — |
case-07 | fail→fail | 18,841 | 61,528 | +227% | 1 | 1 | 0% | 2,950 | 5,886 | +100% | 0 | 0 | — |
case-08 | pass→pass | 20,458 | 26,210 | +28% | 1 | 1 | 0% | 3,083 | 5,330 | +73% | 0 | 0 | — |
case-09 | pass→pass | 23,635 | 23,616 | -0% | 1 | 1 | 0% | 2,812 | 5,536 | +97% | 0 | 0 | — |
case-10 | fail→fail | 27,474 | 38,003 | +38% | 1 | 1 | 0% | 3,652 | 6,744 | +85% | 0 | 0 | — |
case-11 | fail→pass | 21,470 | 43,689 | +103% | 1 | 1 | 0% | 3,506 | 5,527 | +58% | 0 | 0 | — |
case-12 | fail→fail | 26,357 | 32,307 | +23% | 1 | 1 | 0% | 3,604 | 5,817 | +61% | 0 | 0 | — |
case-13 | pass→pass | 22,432 | 27,348 | +22% | 1 | 1 | 0% | 3,003 | 5,265 | +75% | 0 | 0 | — |
case-14 | pass→pass | 18,934 | 30,682 | +62% | 1 | 1 | 0% | 3,269 | 5,457 | +67% | 0 | 0 | — |
case-15 | pass→pass | 18,612 | 31,212 | +68% | 1 | 1 | 0% | 2,890 | 5,551 | +92% | 0 | 0 | — |
case-16 | pass→pass | 20,080 | 26,396 | +31% | 1 | 1 | 0% | 2,699 | 5,040 | +87% | 0 | 0 | — |
case-17 | fail→pass | 19,814 | 22,457 | +13% | 1 | 1 | 0% | 2,900 | 4,798 | +65% | 0 | 0 | — |
case-18 | fail→fail | 22,706 | 30,305 | +33% | 1 | 1 | 0% | 3,028 | 5,372 | +77% | 0 | 0 | — |
case-19 | fail→fail | 28,336 | 29,628 | +5% | 1 | 1 | 0% | 3,928 | 5,774 | +47% | 0 | 0 | — |
case-20 | pass→pass | 54,891 | 27,329 | -50% | 1 | 1 | 0% | 3,472 | 5,671 | +63% | 0 | 0 | — |
case-21 | fail→pass | 20,201 | 17,736 | -12% | 1 | 1 | 0% | 2,635 | 3,725 | +41% | 0 | 0 | — |
case-22 | fail→pass | 12,680 | 18,364 | +45% | 1 | 1 | 0% | 1,534 | 3,693 | +141% | 0 | 0 | — |
case-23 | fail→pass | 16,779 | 21,814 | +30% | 1 | 1 | 0% | 2,213 | 4,141 | +87% | 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 +30 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.