Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →用于合同版本差异识别与风险变更追踪，比较新旧版本条款新增、删除、修改及风险关键词变化。适用于合同谈判轮次管理、法务复核和重大变更提示场景。
.claude/skills/aifinlab-contract-version-compare-assistant/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-06 | ✗→✓ | ▲ Improved | -49% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -38% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -12% | 0% |
| case-01 | ✗→✓ | ▲ Improved | 44% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -47% | 0% |
本技能用于对比合同新旧版本差异，重点识别：
支持 JSON 或 JSONL。推荐输入单个 JSON 对象，包含：
contract_nameold_clauses（条款数组）new_clauses（条款数组）条款对象建议包含：
clause_noclause_titleclause_text使用 scripts/contract_version_compare.py 生成报告：
bashpython scripts/contract_version_compare.py \ --input version_pair.json \ --output version_diff_report.md
可选参数：
--keywords：自定义风险关键词 JSON--top：展示重点变化数量（默认 30）| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-06 | fail→pass | 11,506 | 2,888 | -75% | 1 | 1 | 0% | 1,732 | 886 | -49% | 0 | 0 | — |
case-07 | fail→pass | 10,953 | 4,351 | -60% | 1 | 1 | 0% | 1,915 | 1,187 | -38% | 0 | 0 | — |
case-08 | fail→pass | 14,826 | 9,873 | -33% | 1 | 1 | 0% | 2,546 | 2,240 | -12% | 0 | 0 | — |
case-01 | fail→pass | 15,964 | 30,214 | +89% | 1 | 1 | 0% | 2,527 | 3,630 | +44% | 0 | 0 | — |
case-02 | fail→fail | 12,192 | 18,505 | +52% | 1 | 1 | 0% | 1,896 | 3,376 | +78% | 0 | 0 | — |
case-03 | fail→fail | 11,653 | 5,432 | -53% | 1 | 1 | 0% | 2,301 | 1,414 | -39% | 0 | 0 | — |
case-04 | pass→pass | 14,432 | 9,142 | -37% | 1 | 1 | 0% | 2,415 | 1,242 | -49% | 0 | 0 | — |
case-05 | fail→pass | 8,510 | 2,255 | -74% | 1 | 1 | 0% | 1,359 | 719 | -47% | 0 | 0 | — |
case-09 | fail→pass | 14,112 | 4,899 | -65% | 1 | 1 | 0% | 2,066 | 1,149 | -44% | 0 | 0 | — |
case-10 | fail→pass | 8,391 | 1,842 | -78% | 1 | 1 | 0% | 1,307 | 692 | -47% | 0 | 0 | — |
case-11 | pass→pass | 16,273 | 11,437 | -30% | 1 | 1 | 0% | 2,150 | 2,060 | -4% | 0 | 0 | — |
case-12 | pass→pass | 15,293 | 15,118 | -1% | 1 | 1 | 0% | 2,299 | 2,525 | +10% | 0 | 0 | — |
case-17 | fail→pass | 15,842 | 9,097 | -43% | 1 | 1 | 0% | 2,419 | 1,978 | -18% | 0 | 0 | — |
case-13 | pass→pass | 14,455 | 11,760 | -19% | 1 | 1 | 0% | 2,067 | 2,215 | +7% | 0 | 0 | — |
case-14 | pass→pass | 17,624 | 13,069 | -26% | 1 | 1 | 0% | 2,605 | 2,199 | -16% | 0 | 0 | — |
case-15 | pass→pass | 16,345 | 17,527 | +7% | 1 | 1 | 0% | 2,438 | 3,086 | +27% | 0 | 0 | — |
case-16 | fail→pass | 8,547 | 2,324 | -73% | 1 | 1 | 0% | 1,599 | 872 | -45% | 0 | 0 | — |
case-18 | pass→pass | 6,029 | 2,503 | -58% | 1 | 1 | 0% | 942 | 825 | -12% | 0 | 0 | — |
case-19 | fail→pass | 11,418 | 3,055 | -73% | 1 | 1 | 0% | 1,668 | 850 | -49% | 0 | 0 | — |
case-20 | fail→pass | 11,648 | 19,563 | +68% | 1 | 1 | 0% | 1,717 | 3,371 | +96% | 0 | 0 | — |
case-21 | pass→pass | 30,477 | 32,354 | +6% | 1 | 1 | 0% | 4,864 | 5,642 | +16% | 0 | 0 | — |
case-22 | pass→pass | 18,317 | 19,842 | +8% | 1 | 1 | 0% | 3,034 | 3,488 | +15% | 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 +50 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.