Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →元典法律法规检索技能（开放平台版 https://open.chineselaw.com）。 用于中国法律、行政法规、司法解释、部门规章、地方性法规等规范性文件的检索。 适合：法条语义检索、法条关键词检索、法规级别检索、法条/法规详情查询。 不负责最终法律结论；如需基于检索结果做法律分析，应继续调用 cs-china-lawyer-analyst； 若是"做法律研究备忘录"类完整流程，应交由 legal-research 编排技能调度。
.claude/skills/thomasmoreai-yd-law-search/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-08 | ✗→✓ | ▲ Improved | -35% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 7% | 0% |
| case-11 | ✗→✓ | ▲ Improved | -16% | 0% |
| case-12 | ✗→✓ | ▲ Improved | -15% | 0% |
| case-15 | ✗→✓ | ▲ Improved | 9% | 0% |
检索型 skill，把规范依据稳定、结构化地检索出来。
yd-law-search：检索规范依据（本 skill）yd-case-search：检索案例与类案cs-china-lawyer-analyst：基于检索结果做法律分析legal-research：上层编排技能，按"查法规→查案例→出备忘录"调度上述三者脚本按以下顺序读取 API Key：
YD_OPEN_API_KEY~/.yd_open_api_keyYD_API_KEY~/.yd_api_key.api_key申请入口：<https://open.chineselaw.com/>
ft_detail / fg_detail 配合 id 拉取。--full 开关用于调试或用户明确要求全文展示时使用。默认路由（用于 99% 的"查法律/查规定"提问）：
search（语义法条）search_keyword（关键词法条）ft_detail（法条详情）仅在出现以下显式信号时才走「法规级别」接口：
fg_detail明确禁止：
search_fg 来回答"某规则是什么"——那是 search / search_keyword 的活。search_fg 只回法规清单（标题、发布部门、发布日期），不返回具体条文。bash# 1. 语义检索（自然语言） echo '{"query":"民间借贷利率上限"}' | \ python3 ~/.claude/skills/yd-law-search/scripts/yd_law_search.py search --stdin # 2. 关键词检索（精确） echo '{"keyword":"个人信息 处罚","search_mode":"AND","sxx":"现行有效"}' | \ python3 ~/.claude/skills/yd-law-search/scripts/yd_law_search.py search_keyword --stdin # 3. 法条详情（需法规名 + 中文条号） echo '{"fgmc":"中华人民共和国刑法","ftnum":"第一百条"}' | \ python3 ~/.claude/skills/yd-law-search/scripts/yd_law_search.py ft_detail --stdin # 4. 法规级别检索（仅在用户要"列法规清单"时） echo '{"keyword":"行政处罚","sxx":"现行有效","top_k":8}' | \ python3 ~/.claude/skills/yd-law-search/scripts/yd_law_search.py search_fg --stdin # 5. 法规详情（按 id 或 fgmc） echo '{"fgmc":"中华人民共和国民法典"}' | \ python3 ~/.claude/skills/yd-law-search/scripts/yd_law_search.py fg_detail --stdin # 旧子命令 search_ft_info 仍兼容（自动映射 query→fgmc, ft_name→ftnum） echo '{"query":"民法典","ft_name":"第六百八十条"}' | \ python3 ~/.claude/skills/yd-law-search/scripts/yd_law_search.py search_ft_info --stdin
通用 flag：
--stdin：从 stdin 读 JSON--full：打印全文（默认只打 ≤120 字摘要）--raw：直接打印 API 原始 JSONtop_k = 8；用户明确要更多时再调大。现行有效；查历史版本用 refer_date（仅 ft_detail / fg_detail 支持）。cs-china-lawyer-analyst。| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 15,201 | 5,083 | -67% | 1 | 1 | 0% | 2,875 | 1,528 | -47% | 0 | 0 | — |
case-02 | fail→fail | 8,756 | 5,853 | -33% | 1 | 1 | 0% | 1,623 | 1,550 | -4% | 0 | 0 | — |
case-03 | fail→fail | 12,146 | 7,928 | -35% | 1 | 1 | 0% | 2,134 | 1,710 | -20% | 0 | 0 | — |
case-04 | fail→fail | 21,649 | 7,693 | -64% | 1 | 1 | 0% | 3,204 | 1,733 | -46% | 0 | 0 | — |
case-05 | fail→fail | 24,183 | 8,111 | -66% | 1 | 1 | 0% | 3,559 | 1,811 | -49% | 0 | 0 | — |
case-06 | fail→fail | 30,584 | 28,787 | -6% | 1 | 1 | 0% | 4,638 | 5,720 | +23% | 0 | 0 | — |
case-07 | fail→fail | 16,576 | 6,334 | -62% | 1 | 1 | 0% | 2,364 | 1,580 | -33% | 0 | 0 | — |
case-08 | fail→pass | 17,811 | 4,430 | -75% | 1 | 1 | 0% | 3,135 | 2,026 | -35% | 0 | 0 | — |
case-09 | fail→pass | 11,466 | 4,801 | -58% | 1 | 1 | 0% | 1,962 | 2,091 | +7% | 0 | 0 | — |
case-10 | fail→fail | 14,809 | 7,479 | -49% | 1 | 1 | 0% | 2,741 | 1,683 | -39% | 0 | 0 | — |
case-11 | fail→pass | 12,079 | 3,615 | -70% | 1 | 1 | 0% | 2,219 | 1,858 | -16% | 0 | 0 | — |
case-12 | fail→pass | 14,595 | 6,551 | -55% | 1 | 1 | 0% | 2,520 | 2,133 | -15% | 0 | 0 | — |
case-13 | fail→fail | 7,356 | 7,222 | -2% | 1 | 1 | 0% | 1,358 | 1,671 | +23% | 0 | 0 | — |
case-14 | fail→fail | 11,606 | 5,890 | -49% | 1 | 1 | 0% | 2,026 | 1,659 | -18% | 0 | 0 | — |
case-15 | fail→pass | 11,803 | 6,000 | -49% | 1 | 1 | 0% | 1,961 | 2,146 | +9% | 0 | 0 | — |
case-16 | pass→pass | 12,398 | 3,074 | -75% | 1 | 1 | 0% | 1,836 | 1,712 | -7% | 0 | 0 | — |
case-17 | fail→pass | 11,487 | 2,631 | -77% | 1 | 1 | 0% | 1,975 | 1,569 | -21% | 0 | 0 | — |
case-18 | pass→pass | 12,263 | 7,441 | -39% | 1 | 1 | 0% | 1,869 | 2,354 | +26% | 0 | 0 | — |
case-19 | fail→pass | 11,667 | 1,843 | -84% | 1 | 1 | 0% | 2,065 | 1,584 | -23% | 0 | 0 | — |
case-20 | fail→pass | 10,781 | 1,803 | -83% | 1 | 1 | 0% | 1,873 | 1,529 | -18% | 0 | 0 | — |
case-21 | fail→pass | 13,072 | 7,547 | -42% | 1 | 1 | 0% | 2,306 | 2,805 | +22% | 0 | 0 | — |
case-22 | fail→pass | 11,908 | 2,670 | -78% | 1 | 1 | 0% | 1,752 | 1,558 | -11% | 0 | 0 | — |
case-23 | pass→pass | 14,471 | 3,996 | -72% | 1 | 1 | 0% | 2,358 | 1,865 | -21% | 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. 23 cases were attempted, and 14 counted toward the lift figure. The other 9 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 +43 percentage points is the difference between those two pass rates over the 14 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.