Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →用于信托领域财富与家族信托中的慈善信托匹配助手场景。支持结构化输入处理、规则分析与Markdown结果输出。
.claude/skills/aifinlab-charitable-trust-match/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | 38% | 0% |
| case-01 | ✗→✓ | ▲ Improved | -77% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -49% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -36% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -68% | 0% |
本技能面向“财富与家族信托 / 方案匹配”业务单元，输出结构化分析报告与风险提示，支持尽调、法务、风控及存续管理流程中的基础判断。
id, name, text, status, timestamp 及相关业务字段bashpython scripts/t868_analysis.py --input input.json --output report.md --title "慈善信托匹配助手（T868）"
python# 调用 skill result = run_skill({ "param1": "value1", "param2": "value2" })
bashpython scripts/run_skill.py --input data.json
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-02 | fail→pass | 3,815 | 2,627 | -31% | 1 | 1 | 0% | 663 | 913 | +38% | 0 | 0 | — |
case-01 | fail→pass | 19,125 | 3,427 | -82% | 1 | 1 | 0% | 3,510 | 805 | -77% | 0 | 0 | — |
case-03 | pass→pass | 9,320 | 1,953 | -79% | 1 | 1 | 0% | 1,525 | 788 | -48% | 0 | 0 | — |
case-04 | pass→pass | 9,043 | 2,243 | -75% | 1 | 1 | 0% | 1,488 | 838 | -44% | 0 | 0 | — |
case-05 | pass→pass | 11,823 | 3,296 | -72% | 1 | 1 | 0% | 1,953 | 963 | -51% | 0 | 0 | — |
case-06 | fail→pass | 11,723 | 3,090 | -74% | 1 | 1 | 0% | 1,799 | 926 | -49% | 0 | 0 | — |
case-07 | fail→fail | 11,950 | 1,969 | -84% | 1 | 1 | 0% | 1,888 | 689 | -64% | 0 | 0 | — |
case-08 | pass→pass | 12,330 | 2,097 | -83% | 1 | 1 | 0% | 1,972 | 741 | -62% | 0 | 0 | — |
case-09 | fail→pass | 8,497 | 2,916 | -66% | 1 | 1 | 0% | 1,318 | 849 | -36% | 0 | 0 | — |
case-10 | fail→pass | 14,947 | 2,321 | -84% | 1 | 1 | 0% | 2,415 | 781 | -68% | 0 | 0 | — |
case-11 | pass→pass | 10,699 | 1,763 | -84% | 1 | 1 | 0% | 1,729 | 683 | -60% | 0 | 0 | — |
case-12 | pass→pass | 12,030 | 4,822 | -60% | 1 | 1 | 0% | 1,777 | 1,226 | -31% | 0 | 0 | — |
case-13 | fail→pass | 11,587 | 4,192 | -64% | 1 | 1 | 0% | 1,900 | 1,126 | -41% | 0 | 0 | — |
case-14 | fail→pass | 10,682 | 4,933 | -54% | 1 | 1 | 0% | 1,556 | 1,204 | -23% | 0 | 0 | — |
case-15 | fail→pass | 10,166 | 2,539 | -75% | 1 | 1 | 0% | 1,451 | 844 | -42% | 0 | 0 | — |
case-16 | pass→fail | 10,670 | 3,364 | -68% | 1 | 1 | 0% | 1,572 | 1,010 | -36% | 0 | 0 | — |
case-17 | fail→fail | 9,404 | 4,677 | -50% | 1 | 1 | 0% | 1,411 | 1,194 | -15% | 0 | 0 | — |
case-18 | pass→pass | 6,591 | 3,739 | -43% | 1 | 1 | 0% | 923 | 1,032 | +12% | 0 | 0 | — |
case-19 | pass→pass | 8,978 | 3,462 | -61% | 1 | 1 | 0% | 1,368 | 1,037 | -24% | 0 | 0 | — |
case-20 | pass→pass | 9,743 | 4,561 | -53% | 1 | 1 | 0% | 1,439 | 1,183 | -18% | 0 | 0 | — |
case-21 | pass→pass | 6,692 | 4,894 | -27% | 1 | 1 | 0% | 1,063 | 1,187 | +12% | 0 | 0 | — |
case-22 | fail→pass | 12,099 | 2,975 | -75% | 1 | 1 | 0% | 1,840 | 934 | -49% | 0 | 0 | — |
case-23 | pass→pass | 11,272 | 2,763 | -75% | 1 | 1 | 0% | 1,724 | 927 | -46% | 0 | 0 | — |
case-24 | pass→pass | 6,763 | 2,882 | -57% | 1 | 1 | 0% | 998 | 906 | -9% | 0 | 0 | — |
case-25 | pass→pass | 17,896 | 17,980 | +0% | 1 | 1 | 0% | 2,883 | 3,564 | +24% | 0 | 0 | — |
case-26 | pass→pass | 16,229 | 16,935 | +4% | 1 | 1 | 0% | 2,463 | 2,883 | +17% | 0 | 0 | — |
case-27 | fail→pass | 8,729 | 11,264 | +29% | 1 | 1 | 0% | 1,268 | 2,154 | +70% | 0 | 0 | — |
case-28 | fail→pass | 13,366 | 14,912 | +12% | 1 | 1 | 0% | 2,058 | 2,786 | +35% | 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. 28 cases were attempted. The headline lift of +36 percentage points is the difference between those two pass rates over the 28 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.