Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →当用户需要协助理赔沟通时使用此 skill。适用于理赔申请指导、与保险公司沟通话术、理赔异议处理等场景。
.claude/skills/aifinlab-claims-communication/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -3% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 1% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 38% | 0% |
| case-21 | ✗→✓ | ▲ Improved | 14% | 0% |
| case-05 | ✓→✗ | ▼ Worse | 31% | 0% |
你是一名经验丰富的保险理赔顾问，目标是帮助用户有效与保险公司沟通，提高理赔效率，维护合法权益。
围绕用户的理赔沟通需求，产出结构化、可操作的沟通指南和话术模板。优先帮助用户解决以下任务：
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 28,994 | 39,540 | +36% | 1 | 1 | 0% | 4,336 | 4,194 | -3% | 0 | 0 | — |
case-02 | fail→pass | 32,544 | 28,430 | -13% | 1 | 1 | 0% | 4,574 | 4,633 | +1% | 0 | 0 | — |
case-03 | pass→pass | 13,519 | 14,197 | +5% | 1 | 1 | 0% | 1,896 | 2,596 | +37% | 0 | 0 | — |
case-04 | pass→pass | 16,571 | 17,340 | +5% | 1 | 1 | 0% | 2,297 | 2,981 | +30% | 0 | 0 | — |
case-05 | pass→fail | 21,189 | 26,743 | +26% | 1 | 1 | 0% | 3,108 | 4,071 | +31% | 0 | 0 | — |
case-06 | pass→pass | 25,007 | 23,616 | -6% | 1 | 1 | 0% | 3,412 | 3,806 | +12% | 0 | 0 | — |
case-07 | pass→pass | 18,245 | 23,084 | +27% | 1 | 1 | 0% | 2,766 | 3,691 | +33% | 0 | 0 | — |
case-18 | pass→pass | 21,289 | 23,355 | +10% | 1 | 1 | 0% | 3,047 | 3,944 | +29% | 0 | 0 | — |
case-08 | fail→fail | 19,934 | 19,771 | -1% | 1 | 1 | 0% | 2,822 | 3,368 | +19% | 0 | 0 | — |
case-09 | fail→pass | 16,438 | 19,678 | +20% | 1 | 1 | 0% | 2,350 | 3,254 | +38% | 0 | 0 | — |
case-10 | pass→pass | 21,222 | 20,086 | -5% | 1 | 1 | 0% | 3,116 | 3,364 | +8% | 0 | 0 | — |
case-11 | pass→pass | 20,264 | 21,719 | +7% | 1 | 1 | 0% | 2,947 | 3,636 | +23% | 0 | 0 | — |
case-12 | fail→fail | 21,934 | 20,476 | -7% | 1 | 1 | 0% | 3,297 | 3,622 | +10% | 0 | 0 | — |
case-13 | pass→pass | 19,620 | 23,385 | +19% | 1 | 1 | 0% | 2,915 | 4,112 | +41% | 0 | 0 | — |
case-14 | pass→pass | 19,890 | 20,241 | +2% | 1 | 1 | 0% | 2,763 | 3,556 | +29% | 0 | 0 | — |
case-15 | pass→pass | 23,405 | 20,505 | -12% | 1 | 1 | 0% | 3,194 | 3,378 | +6% | 0 | 0 | — |
case-16 | pass→pass | 11,242 | 14,199 | +26% | 1 | 1 | 0% | 1,634 | 2,647 | +62% | 0 | 0 | — |
case-17 | pass→pass | 19,718 | 19,955 | +1% | 1 | 1 | 0% | 2,663 | 3,376 | +27% | 0 | 0 | — |
case-19 | fail→fail | 30,625 | 29,534 | -4% | 1 | 1 | 0% | 5,836 | 6,087 | +4% | 0 | 0 | — |
case-20 | fail→fail | 26,168 | 24,647 | -6% | 1 | 1 | 0% | 3,972 | 4,448 | +12% | 0 | 0 | — |
case-21 | fail→pass | 22,157 | 23,403 | +6% | 1 | 1 | 0% | 3,456 | 3,955 | +14% | 0 | 0 | — |
case-22 | fail→fail | 15,459 | 20,439 | +32% | 1 | 1 | 0% | 2,483 | 3,455 | +39% | 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 +14 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.