Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use when wealth-management teams need to respond to a redemption request by producing compliant retention talk tracks, options, and follow-up actions based on client goals, liquidity needs, and product facts. Trigger this skill for redemption-risk conversations, retention plans, and client-facing explanations that must avoid收益承诺 and明确风险边界。
.claude/skills/aifinlab-bank-t187-wealth-management-redemption-retention-assistant/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 51% | 0% |
| case-11 | ✗→✓ | ▲ Improved | 85% | 0% |
| case-14 | ✗→✓ | ▲ Improved | 63% | 0% |
| case-17 | ✗→✓ | ▲ Improved | 91% | 0% |
| case-02 | ✓→✓ | ＝ Same ✓ | 81% | 0% |
本技能面向银行财富管理场景，围绕客户赎回请求或赎回风险信号，生成一套可执行的“挽留与沟通动作包”。目标是把客户画像、持仓事实、市场背景与合规边界组织成前台可直接使用的沟通主线、替代方案与后续跟进动作，同时清楚标注信息缺口与禁用表述。
scripts/wealth_redemption_retention.py：根据输入生成挽留动作包（Markdown/JSON）scripts/run_skill.py：命令行入口references/input-schema.md：输入字段口径references/output-schema.md：输出结构口径references/redemption-checklist.md：赎回核验清单references/retention-strategy-framework.md：挽留方案框架references/compliance-red-lines.md：合规禁用表述references/conversation-flow.md：沟通节奏模板assets/templates/retention-plan-template.md：挽留方案模板assets/templates/call-script-template.md：电话/面谈话术模板assets/templates/risk-disclosure-template.md：风险提示模板assets/templates/follow-up-tasklist.md：跟进任务清单模板assets/example-input.json：示例输入| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 25,022 | 30,164 | +21% | 1 | 1 | 0% | 3,113 | 4,716 | +51% | 0 | 0 | — |
case-02 | pass→pass | 18,040 | 27,280 | +51% | 1 | 1 | 0% | 2,219 | 4,024 | +81% | 0 | 0 | — |
case-03 | pass→pass | 22,880 | 24,353 | +6% | 1 | 1 | 0% | 2,463 | 3,982 | +62% | 0 | 0 | — |
case-04 | pass→pass | 20,141 | 26,087 | +30% | 1 | 1 | 0% | 2,880 | 4,723 | +64% | 0 | 0 | — |
case-05 | pass→pass | 22,674 | 23,647 | +4% | 1 | 1 | 0% | 2,819 | 4,697 | +67% | 0 | 0 | — |
case-06 | pass→pass | 17,503 | 24,214 | +38% | 1 | 1 | 0% | 2,734 | 4,146 | +52% | 0 | 0 | — |
case-07 | pass→pass | 25,129 | 32,824 | +31% | 1 | 1 | 0% | 3,166 | 4,910 | +55% | 0 | 0 | — |
case-12 | pass→pass | 22,877 | 30,938 | +35% | 1 | 1 | 0% | 3,018 | 4,943 | +64% | 0 | 0 | — |
case-08 | pass→pass | 28,525 | 20,891 | -27% | 1 | 1 | 0% | 2,378 | 3,843 | +62% | 0 | 0 | — |
case-09 | pass→pass | 24,016 | 22,008 | -8% | 1 | 1 | 0% | 2,932 | 4,004 | +37% | 0 | 0 | — |
case-10 | pass→pass | 29,070 | 30,716 | +6% | 1 | 1 | 0% | 2,814 | 4,981 | +77% | 0 | 0 | — |
case-11 | fail→pass | 20,453 | 25,376 | +24% | 1 | 1 | 0% | 2,609 | 4,832 | +85% | 0 | 0 | — |
case-13 | pass→pass | 21,833 | 26,377 | +21% | 1 | 1 | 0% | 3,039 | 4,310 | +42% | 0 | 0 | — |
case-14 | fail→pass | 21,628 | 28,589 | +32% | 1 | 1 | 0% | 3,314 | 5,406 | +63% | 0 | 0 | — |
case-15 | pass→pass | 21,132 | 29,273 | +39% | 1 | 1 | 0% | 3,002 | 4,671 | +56% | 0 | 0 | — |
case-16 | pass→pass | 22,225 | 25,887 | +16% | 1 | 1 | 0% | 2,720 | 4,721 | +74% | 0 | 0 | — |
case-17 | fail→pass | 19,776 | 30,781 | +56% | 1 | 1 | 0% | 2,818 | 5,391 | +91% | 0 | 0 | — |
case-18 | pass→pass | 19,389 | 21,314 | +10% | 1 | 1 | 0% | 2,267 | 3,700 | +63% | 0 | 0 | — |
case-19 | pass→pass | 22,936 | 26,463 | +15% | 1 | 1 | 0% | 2,658 | 4,548 | +71% | 0 | 0 | — |
case-20 | fail→fail | 22,863 | 39,681 | +74% | 1 | 1 | 0% | 2,852 | 6,865 | +141% | 0 | 0 | — |
case-21 | pass→pass | 23,875 | 34,512 | +45% | 1 | 1 | 0% | 2,486 | 5,332 | +114% | 0 | 0 | — |
case-22 | pass→pass | 21,092 | 17,157 | -19% | 1 | 1 | 0% | 2,492 | 3,570 | +43% | 0 | 0 | — |
case-23 | pass→pass | 21,095 | 40,770 | +93% | 1 | 1 | 0% | 3,014 | 5,179 | +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. 23 cases were attempted. The headline lift of +17 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.