Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →用于信托领域合同与法务中的合同条款助手-补充协议版场景。支持结构化输入处理、规则分析与Markdown结果输出。
.claude/skills/aifinlab-contract-clause-supplemental-agreement/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | -43% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -37% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -21% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -44% | 0% |
| case-12 | ✗→✓ | ▲ Improved | 17% | 0% |
本技能面向“合同与法务 / 条款抽取”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 及相关业务字段按顺序执行：
bashnode scripts/t809_preprocess.js --input input.jsonl --output processed.json
bashpython scripts/t809_analysis.py --input processed.json --rules config/rules.json --baseline reference/baseline.json --output report.md
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-06 | fail→fail | 13,289 | 3,433 | -74% | 1 | 1 | 0% | 1,977 | 1,009 | -49% | 0 | 0 | — |
case-01 | fail→fail | 22,871 | 21,603 | -6% | 1 | 1 | 0% | 3,598 | 3,344 | -7% | 0 | 0 | — |
case-02 | fail→fail | 23,423 | 20,631 | -12% | 1 | 1 | 0% | 3,756 | 3,874 | +3% | 0 | 0 | — |
case-03 | fail→fail | 23,126 | 22,286 | -4% | 1 | 1 | 0% | 3,729 | 4,133 | +11% | 0 | 0 | — |
case-04 | fail→pass | 9,468 | 2,562 | -73% | 1 | 1 | 0% | 1,543 | 879 | -43% | 0 | 0 | — |
case-05 | fail→pass | 10,094 | 2,745 | -73% | 1 | 1 | 0% | 1,607 | 1,007 | -37% | 0 | 0 | — |
case-07 | fail→pass | 15,980 | 8,071 | -49% | 1 | 1 | 0% | 2,234 | 1,758 | -21% | 0 | 0 | — |
case-08 | fail→pass | 10,642 | 2,414 | -77% | 1 | 1 | 0% | 1,532 | 856 | -44% | 0 | 0 | — |
case-09 | pass→fail | 17,510 | 13,619 | -22% | 1 | 1 | 0% | 2,986 | 2,845 | -5% | 0 | 0 | — |
case-10 | fail→fail | 19,532 | 19,546 | +0% | 1 | 1 | 0% | 2,941 | 3,477 | +18% | 0 | 0 | — |
case-11 | fail→fail | 19,750 | 17,763 | -10% | 1 | 1 | 0% | 2,932 | 3,203 | +9% | 0 | 0 | — |
case-12 | fail→pass | 14,907 | 14,468 | -3% | 1 | 1 | 0% | 2,151 | 2,524 | +17% | 0 | 0 | — |
case-13 | pass→pass | 17,550 | 16,390 | -7% | 1 | 1 | 0% | 2,633 | 2,912 | +11% | 0 | 0 | — |
case-14 | fail→pass | 17,472 | 11,775 | -33% | 1 | 1 | 0% | 2,487 | 2,165 | -13% | 0 | 0 | — |
case-15 | fail→pass | 6,731 | 2,663 | -60% | 1 | 1 | 0% | 1,125 | 958 | -15% | 0 | 0 | — |
case-16 | pass→pass | 16,034 | 5,431 | -66% | 1 | 1 | 0% | 2,516 | 1,034 | -59% | 0 | 0 | — |
case-17 | fail→pass | 25,781 | 2,476 | -90% | 1 | 1 | 0% | 1,741 | 900 | -48% | 0 | 0 | — |
case-18 | pass→pass | 9,510 | 2,356 | -75% | 1 | 1 | 0% | 1,504 | 845 | -44% | 0 | 0 | — |
case-19 | pass→pass | 14,970 | 12,516 | -16% | 1 | 1 | 0% | 2,290 | 2,457 | +7% | 0 | 0 | — |
case-20 | pass→pass | 11,700 | 2,988 | -74% | 1 | 1 | 0% | 1,620 | 931 | -43% | 0 | 0 | — |
case-21 | pass→pass | 12,511 | 3,399 | -73% | 1 | 1 | 0% | 1,815 | 1,046 | -42% | 0 | 0 | — |
case-22 | pass→pass | 14,272 | 3,832 | -73% | 1 | 1 | 0% | 2,390 | 1,152 | -52% | 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, and 21 counted toward the lift figure. The other 1 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +32 percentage points is the difference between those two pass rates over the 21 comparable cases. 2 cases got worse with the skill loaded, and they are 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.