Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →当需要在银行财富管理场景下，为客户提供市场波动安抚、情绪稳定与理性陪伴的合规沟通方案时使用本技能；适合输出可直接用于一线沟通的主线话术、安抚结构、禁用表述提醒与后续跟进行动。
.claude/skills/aifinlab-bank-t186-wealth-management-volatility-comfort-assistant/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | 79% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 77% | 0% |
| case-20 | ✗→✓ | ▲ Improved | 26% | 0% |
| case-22 | ✗→✓ | ▲ Improved | 28% | 0% |
| case-17 | ✗→✓ | ▲ Improved | 72% | 0% |
在市场波动或净值回撤背景下，基于客户画像、风险承受能力、持仓结构与事件背景，输出合规、克制、可复用的安抚沟通方案。强调“安抚≠承诺”，通过事实澄清、情绪共情、策略解释和下一步动作，帮助一线人员稳定客户预期、避免误导式表达，并形成可留痕的沟通材料。
references/input-schema.md：输入字段规范references/output-schema.md：输出结构规范references/comfort-playbook.md：安抚主线与分层沟通策略references/compliance-do-dont.md：合规禁用表述与替代表达references/empathy-ladder.md：情绪分层与共情梯度references/scenario-matrix.md：客户类型 × 市场情绪的沟通矩阵assets/example-input.json：示例输入assets/templates/comfort-talk-track.md：安抚话术模板assets/templates/follow-up-plan.md：跟进计划模板scripts/volatility_comfort_pack.py：生成安抚话术与跟进行动包scripts/run_skill.py：脚本入口| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 29,865 | 69,893 | +134% | 1 | 1 | 0% | 3,906 | 5,667 | +45% | 0 | 0 | — |
case-02 | fail→fail | 22,737 | 27,530 | +21% | 1 | 1 | 0% | 3,234 | 4,582 | +42% | 0 | 0 | — |
case-03 | fail→fail | 23,599 | 20,696 | -12% | 1 | 1 | 0% | 2,733 | 3,903 | +43% | 0 | 0 | — |
case-04 | fail→pass | 21,549 | 24,959 | +16% | 1 | 1 | 0% | 2,623 | 4,701 | +79% | 0 | 0 | — |
case-05 | pass→fail | 24,185 | 26,754 | +11% | 1 | 1 | 0% | 3,048 | 4,731 | +55% | 0 | 0 | — |
case-06 | fail→pass | 15,181 | 20,015 | +32% | 1 | 1 | 0% | 2,152 | 3,816 | +77% | 0 | 0 | — |
case-07 | pass→pass | 20,097 | 24,673 | +23% | 1 | 1 | 0% | 2,934 | 4,607 | +57% | 0 | 0 | — |
case-08 | fail→fail | 25,227 | 27,021 | +7% | 1 | 1 | 0% | 2,930 | 4,444 | +52% | 0 | 0 | — |
case-20 | fail→pass | 23,746 | 23,048 | -3% | 1 | 1 | 0% | 3,310 | 4,155 | +26% | 0 | 0 | — |
case-21 | pass→pass | 20,923 | 25,289 | +21% | 1 | 1 | 0% | 2,535 | 4,080 | +61% | 0 | 0 | — |
case-22 | fail→pass | 26,362 | 31,828 | +21% | 1 | 1 | 0% | 3,725 | 4,763 | +28% | 0 | 0 | — |
case-09 | pass→pass | 36,865 | 33,658 | -9% | 1 | 1 | 0% | 3,093 | 5,265 | +70% | 0 | 0 | — |
case-10 | pass→pass | 22,789 | 28,121 | +23% | 1 | 1 | 0% | 2,737 | 4,517 | +65% | 0 | 0 | — |
case-11 | fail→fail | 21,027 | 25,321 | +20% | 1 | 1 | 0% | 2,828 | 4,678 | +65% | 0 | 0 | — |
case-12 | pass→pass | 23,962 | 30,819 | +29% | 1 | 1 | 0% | 2,993 | 5,319 | +78% | 0 | 0 | — |
case-13 | pass→pass | 29,288 | 34,687 | +18% | 1 | 1 | 0% | 3,512 | 5,384 | +53% | 0 | 0 | — |
case-14 | pass→pass | 23,024 | 35,056 | +52% | 1 | 1 | 0% | 3,185 | 4,846 | +52% | 0 | 0 | — |
case-15 | pass→pass | 21,033 | 24,662 | +17% | 1 | 1 | 0% | 2,689 | 4,242 | +58% | 0 | 0 | — |
case-16 | pass→pass | 22,076 | 28,400 | +29% | 1 | 1 | 0% | 2,929 | 4,669 | +59% | 0 | 0 | — |
case-17 | fail→pass | 19,870 | 29,696 | +49% | 1 | 1 | 0% | 2,780 | 4,774 | +72% | 0 | 0 | — |
case-18 | fail→fail | 22,794 | 20,823 | -9% | 1 | 1 | 0% | 2,820 | 4,101 | +45% | 0 | 0 | — |
case-19 | pass→pass | 26,725 | 31,733 | +19% | 1 | 1 | 0% | 3,161 | 5,121 | +62% | 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 +18 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.