Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →当用户需要在银行财富管理场景下，对客户持仓波动进行成因拆解、指标解读、驱动项归因并输出可沟通的解释与后续动作建议时使用本技能。适合在持仓波动沟通、客户陪伴或阶段性诊断场景中触发。
.claude/skills/aifinlab-bank-t185-wealth-management-volatility-explanation-assistant/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 40% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 3% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 1% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -2% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 41% | 0% |
本技能用于在银行财富管理场景中，围绕客户持仓波动进行结构化解释与沟通输出。目标是把“波动事实、成因拆解、风险提示、后续动作”拆开讲清楚，并在信息不足时给出清晰的补充清单与风险边界。
当需要稳定复用的归因口径、结构化报告或批量处理时，使用脚本输出标准化结果。
功能：
输入格式（JSON 示例）：
json{ "client": { "name": "张三", "risk_profile": "稳健", "investment_goal": "稳健增值", "horizon": "3-5 年", "liquidity_needs": "中等" }, "portfolio": { "as_of_date": "2026-03-01", "period": "2026-02-01 ~ 2026-02-29", "total_value": 1200000, "base_currency": "CNY", "positions": [ {"name": "权益A", "asset_class": "权益", "weight": 0.35, "return_pct": -0.042}, {"name": "固收B", "asset_class": "固收", "weight": 0.45, "return_pct": 0.006}, {"name": "另类C", "asset_class": "另类", "weight": 0.20, "return_pct": -0.015} ] }, "benchmark": {"name": "稳健组合基准", "return_pct": -0.018}, "market": { "summary": "权益市场波动放大，利率曲线小幅下行", "key_events": ["权益板块估值回调", "利率中枢下移"] } }
命令行示例：
bashpython scripts/volatility_explainer.py --input input.json --format markdown --output report.md
输出说明：
markdown：用于客户沟通或内部汇报的结构化说明json：用于后续系统处理或二次加工功能：
volatility_explainer.py 的入口脚本| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 26,041 | 59,766 | +130% | 1 | 1 | 0% | 4,116 | 5,779 | +40% | 0 | 0 | — |
case-02 | fail→pass | 10,117 | 5,772 | -43% | 1 | 1 | 0% | 1,470 | 1,507 | +3% | 0 | 0 | — |
case-03 | pass→pass | 12,686 | 2,351 | -81% | 1 | 1 | 0% | 1,702 | 1,614 | -5% | 0 | 0 | — |
case-04 | fail→pass | 10,778 | 4,520 | -58% | 1 | 1 | 0% | 1,829 | 1,847 | +1% | 0 | 0 | — |
case-14 | pass→pass | 8,969 | 6,200 | -31% | 1 | 1 | 0% | 1,393 | 2,237 | +61% | 0 | 0 | — |
case-05 | fail→pass | 9,834 | 3,472 | -65% | 1 | 1 | 0% | 1,653 | 1,616 | -2% | 0 | 0 | — |
case-06 | fail→pass | 9,488 | 7,062 | -26% | 1 | 1 | 0% | 1,393 | 1,964 | +41% | 0 | 0 | — |
case-07 | pass→pass | 12,243 | 10,737 | -12% | 1 | 1 | 0% | 1,624 | 2,602 | +60% | 0 | 0 | — |
case-08 | pass→pass | 12,709 | 17,090 | +34% | 1 | 1 | 0% | 1,745 | 3,667 | +110% | 0 | 0 | — |
case-09 | fail→pass | 11,280 | 14,021 | +24% | 1 | 1 | 0% | 1,516 | 2,831 | +87% | 0 | 0 | — |
case-10 | pass→pass | 14,101 | 21,561 | +53% | 1 | 1 | 0% | 2,050 | 3,808 | +86% | 0 | 0 | — |
case-11 | fail→fail | 17,848 | 22,412 | +26% | 1 | 1 | 0% | 3,124 | 4,704 | +51% | 0 | 0 | — |
case-12 | pass→pass | 10,815 | 18,709 | +73% | 1 | 1 | 0% | 1,685 | 3,828 | +127% | 0 | 0 | — |
case-13 | fail→pass | 14,758 | 20,636 | +40% | 1 | 1 | 0% | 2,168 | 3,743 | +73% | 0 | 0 | — |
case-15 | pass→pass | 3,920 | 6,708 | +71% | 1 | 1 | 0% | 731 | 2,344 | +221% | 0 | 0 | — |
case-16 | pass→pass | 12,399 | 18,006 | +45% | 1 | 1 | 0% | 2,027 | 4,139 | +104% | 0 | 0 | — |
case-17 | fail→pass | 13,420 | 8,072 | -40% | 1 | 1 | 0% | 1,904 | 2,396 | +26% | 0 | 0 | — |
case-18 | pass→pass | 15,225 | 3,396 | -78% | 1 | 1 | 0% | 2,094 | 1,705 | -19% | 0 | 0 | — |
case-19 | fail→pass | 9,634 | 15,723 | +63% | 1 | 1 | 0% | 1,674 | 3,350 | +100% | 0 | 0 | — |
case-20 | fail→pass | 12,262 | 9,260 | -24% | 1 | 1 | 0% | 1,607 | 2,358 | +47% | 0 | 0 | — |
case-21 | pass→pass | 11,989 | 19,107 | +59% | 1 | 1 | 0% | 1,835 | 3,642 | +98% | 0 | 0 | — |
case-22 | fail→pass | 18,662 | 2,495 | -87% | 1 | 1 | 0% | 2,146 | 1,477 | -31% | 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 +50 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.