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，围绕路演支持生成客户沟通、陪伴内容、路演材料或合规表达时使用本技能，适合输出可直接使用的话术主线、内容结构、禁用表述提醒和后续跟进建议。
.claude/skills/aifinlab-bank-t191-wealth-management-roadshow-support-assistant/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-09 | ✗→✓ | ▲ Improved | 67% | 0% |
| case-22 | ✗→✓ | ▲ Improved | 46% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -40% | 0% |
| case-13 | ✗→✓ | ▲ Improved | 13% | 0% |
| case-14 | ✗→✓ | ▲ Improved | 24% | 0% |
本技能面向银行财富管理场景，核心是把客户目标、风险承受能力、期限约束、家庭资产配置和市场环境组织成可解释、可沟通、可合规落地的路演支持材料。主要服务理财经理、财富顾问和私行团队，输出时要清晰区分已确认事实、待核验信息与经验判断。
当需要把结构化输入快速转成标准化路演稿与风险提示清单时，使用 scripts/roadshow_support_builder.py：
输入 JSON（示例字段）
client.profile: 客户层级、风险偏好、期限、流动性要求client.goal: 本次路演的核心目标与关注点products: 产品列表（名称、类型、卖点、风险等级、适当性要求）market_view: 近期观点与来源、有效期compliance: 禁用表述、必选风险提示、渠道限制follow_up: 建议的跟进动作与节奏输出 Markdown
路演主线观点与配置逻辑产品要点风险提示与禁用表述常见追问与回应后续跟进| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 34,717 | 41,214 | +19% | 1 | 1 | 0% | 4,866 | 6,423 | +32% | 0 | 0 | — |
case-02 | fail→fail | 37,022 | 32,575 | -12% | 1 | 1 | 0% | 4,642 | 5,850 | +26% | 0 | 0 | — |
case-03 | fail→fail | 33,058 | 38,634 | +17% | 1 | 1 | 0% | 4,457 | 5,645 | +27% | 0 | 0 | — |
case-04 | pass→pass | 10,194 | 15,114 | +48% | 1 | 1 | 0% | 1,402 | 3,306 | +136% | 0 | 0 | — |
case-09 | fail→pass | 18,952 | 23,387 | +23% | 1 | 1 | 0% | 2,645 | 4,426 | +67% | 0 | 0 | — |
case-10 | pass→pass | 27,283 | 27,118 | -1% | 1 | 1 | 0% | 3,328 | 4,892 | +47% | 0 | 0 | — |
case-21 | pass→pass | 7,936 | 4,314 | -46% | 1 | 1 | 0% | 1,220 | 1,589 | +30% | 0 | 0 | — |
case-22 | fail→pass | 22,136 | 25,408 | +15% | 1 | 1 | 0% | 2,776 | 4,051 | +46% | 0 | 0 | — |
case-05 | pass→pass | 22,750 | 16,239 | -29% | 1 | 1 | 0% | 2,534 | 3,339 | +32% | 0 | 0 | — |
case-06 | pass→pass | 20,233 | 15,739 | -22% | 1 | 1 | 0% | 2,498 | 3,176 | +27% | 0 | 0 | — |
case-07 | fail→pass | 26,449 | 7,172 | -73% | 1 | 1 | 0% | 3,430 | 2,074 | -40% | 0 | 0 | — |
case-08 | fail→fail | 30,028 | 25,128 | -16% | 1 | 1 | 0% | 3,568 | 4,611 | +29% | 0 | 0 | — |
case-11 | pass→pass | 23,254 | 27,276 | +17% | 1 | 1 | 0% | 2,574 | 4,689 | +82% | 0 | 0 | — |
case-12 | pass→pass | 18,825 | 25,777 | +37% | 1 | 1 | 0% | 2,843 | 4,335 | +52% | 0 | 0 | — |
case-13 | fail→pass | 16,450 | 14,295 | -13% | 1 | 1 | 0% | 2,573 | 2,897 | +13% | 0 | 0 | — |
case-14 | fail→pass | 25,222 | 22,158 | -12% | 1 | 1 | 0% | 3,405 | 4,223 | +24% | 0 | 0 | — |
case-20 | fail→fail | 19,716 | 27,193 | +38% | 1 | 1 | 0% | 2,693 | 4,320 | +60% | 0 | 0 | — |
case-15 | fail→fail | 30,386 | 25,940 | -15% | 1 | 1 | 0% | 3,859 | 4,936 | +28% | 0 | 0 | — |
case-16 | pass→pass | 25,445 | 28,424 | +12% | 1 | 1 | 0% | 3,416 | 4,915 | +44% | 0 | 0 | — |
case-17 | fail→pass | 13,270 | 4,001 | -70% | 1 | 1 | 0% | 1,914 | 1,553 | -19% | 0 | 0 | — |
case-18 | fail→fail | 27,221 | 39,979 | +47% | 1 | 1 | 0% | 3,096 | 4,492 | +45% | 0 | 0 | — |
case-19 | fail→pass | 27,172 | 32,091 | +18% | 1 | 1 | 0% | 2,945 | 5,487 | +86% | 0 | 0 | — |
case-23 | pass→pass | 16,100 | 30,244 | +88% | 1 | 1 | 0% | 2,094 | 3,716 | +77% | 0 | 0 | — |
case-24 | fail→fail | 33,890 | 28,202 | -17% | 1 | 1 | 0% | 3,202 | 4,696 | +47% | 0 | 0 | — |
case-25 | fail→pass | 11,412 | 5,263 | -54% | 1 | 1 | 0% | 1,818 | 1,401 | -23% | 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. 25 cases were attempted. The headline lift of +32 percentage points is the difference between those two pass rates over the 25 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.