Install any skill in seconds. Free to start, no credit card required.
Get Started Free →当用户需要判断保险理赔案件中的重大疾病场景是否可能落入保险责任范围、分析某个重大疾病是否达到赔付定义、结合保单和条款判断重疾责任是否可能触发、识别等待期既往症除外责任定义边界争议或评估材料是否足以支持重疾责任判断时使用本 skill。适用于分析理赔系统案件记录、报案信息、门诊或住院病历、诊断证明、病理报告、影像检查报告、实验室检查报告、手术记录、出院小结、会诊意见、基因检测结果、专科医生说明、客服或审核备注、OCR 文本、PDF 文档和截图转写内容,并结合产品责任、重疾定义、等待期、责任免除、轻症中症重疾边界、多次赔付限制和疾病状态要求形成结构化初步判断。
.claude/skills/aifinlab-critical-illness-coverage-scope-judgment-assistant/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-22 | ✗→✓ | ▲ Improved | 73% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 86% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 163% | 0% |
| case-11 | ✗→✓ | ▲ Improved | 180% | 0% |
| case-15 | ✗→✓ | ▲ Improved | 63% | 0% |
你是“责任范围判断助手-重疾版”。
围绕“辅助判断重疾场景是否可能落入责任范围”开展分析,重点整理案件事实、疾病定义匹配、责任触发条件、免责限制、材料充分性和复核建议,帮助理赔、复核、客服支持和运营人员更快形成一致、可追溯的重疾责任判断意见。
只做重疾场景的责任范围辅助判断、定义匹配分析和复核支持,不替代理赔、复核、法务或医疗审核岗位作出正式赔付决定。
coverage-scope-judgment:适用于更通用的责任范围判断。claims-material-check-critical-illness-assistant:适用于重疾理赔材料完整性检查,不直接输出定义匹配和责任边界判断。outpatient-coverage-scope-judgment-assistant:适用于门诊责任范围判断,不适用于重疾定义匹配。以下情况优先使用本技能:
以下情况不按本技能直接处理,应提示这是更细分的问题:
写作时遵循以下规则:
输出应让理赔、复核、客服支持或运营人员能快速回答:
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-22 | fail→pass | 14,391 | 12,604 | -12% | 1 | 1 | 0% | 2,471 | 4,268 | +73% | 0 | 0 | — |
case-01 | fail→fail | 25,443 | 31,703 | +25% | 1 | 1 | 0% | 4,000 | 7,102 | +78% | 0 | 0 | — |
case-02 | fail→pass | 24,982 | 28,713 | +15% | 1 | 1 | 0% | 3,765 | 7,017 | +86% | 0 | 0 | — |
case-03 | fail→fail | 32,746 | 33,630 | +3% | 1 | 1 | 0% | 4,712 | 7,723 | +64% | 0 | 0 | — |
case-04 | pass→pass | 13,146 | 29,100 | +121% | 1 | 1 | 0% | 2,214 | 7,169 | +224% | 0 | 0 | — |
case-05 | pass→pass | 20,007 | 25,345 | +27% | 1 | 1 | 0% | 3,176 | 6,461 | +103% | 0 | 0 | — |
case-06 | pass→pass | 17,747 | 24,418 | +38% | 1 | 1 | 0% | 2,832 | 6,451 | +128% | 0 | 0 | — |
case-07 | fail→fail | 17,508 | 28,296 | +62% | 1 | 1 | 0% | 2,663 | 6,474 | +143% | 0 | 0 | — |
case-08 | pass→pass | 23,275 | 30,781 | +32% | 1 | 1 | 0% | 3,416 | 7,405 | +117% | 0 | 0 | — |
case-09 | fail→pass | 15,177 | 22,846 | +51% | 1 | 1 | 0% | 2,245 | 5,894 | +163% | 0 | 0 | — |
case-10 | pass→pass | 15,897 | 24,542 | +54% | 1 | 1 | 0% | 2,504 | 6,214 | +148% | 0 | 0 | — |
case-11 | fail→pass | 12,855 | 20,458 | +59% | 1 | 1 | 0% | 2,092 | 5,867 | +180% | 0 | 0 | — |
case-12 | pass→fail | 17,889 | 21,554 | +20% | 1 | 1 | 0% | 2,530 | 5,716 | +126% | 0 | 0 | — |
case-13 | pass→pass | 24,263 | 27,243 | +12% | 1 | 1 | 0% | 3,676 | 6,606 | +80% | 0 | 0 | — |
case-14 | fail→fail | 14,750 | 20,959 | +42% | 1 | 1 | 0% | 2,149 | 5,654 | +163% | 0 | 0 | — |
case-15 | fail→pass | 18,387 | 16,300 | -11% | 1 | 1 | 0% | 3,029 | 4,945 | +63% | 0 | 0 | — |
case-16 | fail→pass | 10,773 | 18,207 | +69% | 1 | 1 | 0% | 1,765 | 5,255 | +198% | 0 | 0 | — |
case-17 | pass→pass | 19,789 | 25,454 | +29% | 1 | 1 | 0% | 2,992 | 6,302 | +111% | 0 | 0 | — |
case-18 | pass→pass | 17,213 | 18,402 | +7% | 1 | 1 | 0% | 2,552 | 5,526 | +117% | 0 | 0 | — |
case-19 | fail→pass | 14,128 | 19,891 | +41% | 1 | 1 | 0% | 2,107 | 5,539 | +163% | 0 | 0 | — |
case-20 | pass→pass | 21,942 | 25,259 | +15% | 1 | 1 | 0% | 3,451 | 6,885 | +100% | 0 | 0 | — |
case-21 | fail→pass | 16,285 | 6,038 | -63% | 1 | 1 | 0% | 2,623 | 3,364 | +28% | 0 | 0 | — |
case-23 | fail→pass | 19,580 | 22,556 | +15% | 1 | 1 | 0% | 2,848 | 6,100 | +114% | 0 | 0 | — |
case-24 | fail→pass | 30,013 | 9,883 | -67% | 1 | 1 | 0% | 4,231 | 3,935 | -7% | 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. 24 cases were attempted. The headline lift of +38 percentage points is the difference between those two pass rates over the 24 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.