Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →当用户需要在银行零售金融场景下，为具体客诉生成安抚沟通话术、表达结构与跟进建议时使用本技能。适合输出可直接使用的沟通主线、禁用表述提醒与跟进计划。
.claude/skills/aifinlab-bank-t169-retail-finance-complaint-comfort-assistant/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-11 | ✗→✓ | ▲ Improved | 14% | 0% |
| case-12 | ✗→✓ | ▲ Improved | 9% | 0% |
| case-15 | ✗→✓ | ▲ Improved | 1% | 0% |
| case-16 | ✗→✓ | ▲ Improved | 13% | 0% |
| case-18 | ✗→✓ | ▲ Improved | 32% | 0% |
本技能用于在合规边界内生成可直接使用的沟通主线，帮助一线人员稳定情绪、澄清事实、说明处理路径。强调“先安抚、再解释、后跟进”，不做收益承诺。
casereferences/input-schema.mdbashpython scripts/run_skill.py --input assets/example-input.json --format markdown python scripts/run_skill.py --input assets/example-input.json --format json
脚本入口：scripts/run_skill.py（调用 shared/retail_service_ops_skill_engine.py 的 t169 场景）。
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 26,111 | 26,836 | +3% | 1 | 1 | 0% | 3,220 | 3,595 | +12% | 0 | 0 | — |
case-02 | fail→fail | 26,403 | 23,670 | -10% | 1 | 1 | 0% | 2,915 | 3,345 | +15% | 0 | 0 | — |
case-03 | fail→fail | 23,258 | 22,895 | -2% | 1 | 1 | 0% | 3,100 | 3,382 | +9% | 0 | 0 | — |
case-04 | pass→pass | 24,578 | 23,685 | -4% | 1 | 1 | 0% | 2,968 | 3,453 | +16% | 0 | 0 | — |
case-05 | fail→fail | 20,900 | 24,610 | +18% | 1 | 1 | 0% | 2,868 | 3,509 | +22% | 0 | 0 | — |
case-06 | fail→fail | 26,500 | 29,492 | +11% | 1 | 1 | 0% | 3,484 | 3,914 | +12% | 0 | 0 | — |
case-07 | pass→pass | 26,137 | 23,215 | -11% | 1 | 1 | 0% | 3,156 | 3,470 | +10% | 0 | 0 | — |
case-08 | fail→fail | 24,019 | 17,658 | -26% | 1 | 1 | 0% | 2,980 | 2,902 | -3% | 0 | 0 | — |
case-09 | fail→fail | 21,017 | 25,650 | +22% | 1 | 1 | 0% | 3,260 | 3,427 | +5% | 0 | 0 | — |
case-10 | pass→pass | 24,577 | 23,690 | -4% | 1 | 1 | 0% | 3,375 | 3,509 | +4% | 0 | 0 | — |
case-11 | fail→pass | 20,605 | 18,535 | -10% | 1 | 1 | 0% | 2,788 | 3,190 | +14% | 0 | 0 | — |
case-12 | fail→pass | 20,086 | 22,198 | +11% | 1 | 1 | 0% | 2,965 | 3,222 | +9% | 0 | 0 | — |
case-13 | pass→pass | 27,369 | 23,795 | -13% | 1 | 1 | 0% | 3,289 | 3,591 | +9% | 0 | 0 | — |
case-14 | pass→pass | 28,953 | 24,774 | -14% | 1 | 1 | 0% | 3,633 | 3,611 | -1% | 0 | 0 | — |
case-15 | fail→pass | 34,169 | 22,653 | -34% | 1 | 1 | 0% | 3,254 | 3,277 | +1% | 0 | 0 | — |
case-16 | fail→pass | 25,868 | 20,147 | -22% | 1 | 1 | 0% | 3,176 | 3,586 | +13% | 0 | 0 | — |
case-17 | fail→fail | 25,663 | 22,269 | -13% | 1 | 1 | 0% | 3,239 | 3,262 | +1% | 0 | 0 | — |
case-18 | fail→pass | 21,790 | 24,807 | +14% | 1 | 1 | 0% | 3,061 | 4,029 | +32% | 0 | 0 | — |
case-19 | fail→pass | 22,836 | 24,124 | +6% | 1 | 1 | 0% | 3,190 | 3,537 | +11% | 0 | 0 | — |
case-20 | pass→pass | 20,789 | 20,215 | -3% | 1 | 1 | 0% | 2,829 | 3,471 | +23% | 0 | 0 | — |
case-21 | pass→pass | 18,471 | 17,077 | -8% | 1 | 1 | 0% | 2,619 | 2,998 | +14% | 0 | 0 | — |
case-22 | pass→pass | 14,027 | 11,110 | -21% | 1 | 1 | 0% | 1,658 | 1,928 | +16% | 0 | 0 | — |
case-23 | fail→fail | 30,204 | 24,566 | -19% | 1 | 1 | 0% | 3,658 | 4,370 | +19% | 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 +26 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.