Install any skill in seconds. Free to start, no credit card required.
Get Started Free →当用户需要查找与理赔案件相关的保险条款、定位责任定义和免责条文、根据案件事实匹配相关保险条款、提取影响理赔判断的关键条文或输出适合理赔审核、复核、客服支持和调查使用的理赔条款定位结果时使用本 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、重疾身故事故相关资料、费用清单与发票、调查记录、客服或审核备注、OCR 文本、PDF 文档、截图转写内容以及产品条款全文、特别约定、责任免除条款、等待期约定、多次赔付规则、医院范围和条款版本信息,并形成结构化条款索引与摘要。
.claude/skills/aifinlab-claims-clause-locator-assistant/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-03 | ✗→✓ | ▲ Improved | 59% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 38% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 101% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 83% | 0% |
| case-11 | ✗→✓ | ▲ Improved | 155% | 0% |
你是“理赔条款定位助手”。
围绕“快速定位与案件相关的关键条款”开展分析,重点整理案件事实、责任条款、定义条款、免责条款、边界限制条款和特别约定的关联关系,帮助理赔、复核、客服支持和调查人员缩短条款查找时间,提升案件处理效率与口径一致性。
只做理赔条款定位、条款关联梳理和复核支持,不替代理赔、复核、法务或调查岗位作出正式责任结论或拒赔决定。
coverage-scope-judgment:适用于更通用的责任范围判断。claims-exclusion-identification-assistant:适用于免责条款识别和除外风险分析,不以完整条款定位为核心。outpatient-coverage-scope-judgment-assistant、critical-illness-coverage-scope-judgment-assistant、death-coverage-scope-judgment-assistant:适用于特定责任场景判断,不以跨类条款定位为核心。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、调查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-01 | fail→fail | 29,154 | 37,513 | +29% | 1 | 1 | 0% | 4,576 | 7,585 | +66% | 0 | 0 | — |
case-02 | fail→fail | 15,668 | 21,384 | +36% | 1 | 1 | 0% | 3,021 | 5,916 | +96% | 0 | 0 | — |
case-03 | fail→pass | 24,626 | 22,657 | -8% | 1 | 1 | 0% | 3,602 | 5,739 | +59% | 0 | 0 | — |
case-04 | fail→pass | 24,541 | 17,997 | -27% | 1 | 1 | 0% | 3,343 | 4,614 | +38% | 0 | 0 | — |
case-05 | fail→pass | 19,726 | 25,406 | +29% | 1 | 1 | 0% | 3,049 | 6,115 | +101% | 0 | 0 | — |
case-06 | pass→pass | 17,351 | 22,794 | +31% | 1 | 1 | 0% | 2,541 | 5,633 | +122% | 0 | 0 | — |
case-07 | pass→pass | 18,966 | 27,302 | +44% | 1 | 1 | 0% | 2,966 | 6,560 | +121% | 0 | 0 | — |
case-08 | pass→pass | 19,516 | 24,623 | +26% | 1 | 1 | 0% | 2,928 | 6,091 | +108% | 0 | 0 | — |
case-09 | pass→pass | 17,871 | 21,933 | +23% | 1 | 1 | 0% | 2,827 | 5,835 | +106% | 0 | 0 | — |
case-10 | fail→pass | 20,126 | 21,385 | +6% | 1 | 1 | 0% | 3,035 | 5,555 | +83% | 0 | 0 | — |
case-11 | fail→pass | 12,993 | 22,165 | +71% | 1 | 1 | 0% | 2,180 | 5,566 | +155% | 0 | 0 | — |
case-12 | pass→pass | 17,864 | 21,895 | +23% | 1 | 1 | 0% | 2,810 | 5,725 | +104% | 0 | 0 | — |
case-13 | fail→pass | 14,266 | 21,693 | +52% | 1 | 1 | 0% | 2,292 | 5,540 | +142% | 0 | 0 | — |
case-14 | pass→pass | 17,116 | 29,405 | +72% | 1 | 1 | 0% | 2,588 | 6,782 | +162% | 0 | 0 | — |
case-15 | pass→pass | 18,267 | 24,951 | +37% | 1 | 1 | 0% | 2,608 | 6,042 | +132% | 0 | 0 | — |
case-16 | pass→pass | 14,121 | 22,477 | +59% | 1 | 1 | 0% | 2,165 | 5,803 | +168% | 0 | 0 | — |
case-17 | fail→pass | 15,866 | 21,955 | +38% | 1 | 1 | 0% | 2,524 | 5,591 | +122% | 0 | 0 | — |
case-18 | pass→pass | 18,171 | 22,055 | +21% | 1 | 1 | 0% | 2,575 | 5,733 | +123% | 0 | 0 | — |
case-19 | pass→pass | 14,934 | 22,832 | +53% | 1 | 1 | 0% | 2,317 | 5,621 | +143% | 0 | 0 | — |
case-20 | pass→pass | 15,732 | 22,768 | +45% | 1 | 1 | 0% | 2,334 | 5,788 | +148% | 0 | 0 | — |
case-21 | pass→pass | 17,279 | 25,521 | +48% | 1 | 1 | 0% | 2,548 | 6,143 | +141% | 0 | 0 | — |
case-22 | fail→pass | 22,206 | 23,518 | +6% | 1 | 1 | 0% | 3,533 | 5,819 | +65% | 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 +36 percentage points is the difference between those two pass rates over the 22 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.