Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →将法条/规范文件（.txt/.docx/.pdf）转为 Markdown。适用于用户要求“法条转 markdown”“pdf/docx 转 markdown”。处理 .pdf/.docx 时先检查是否已安装 mineru-ocr skill；未安装先引导安装，安装后优先用 mineru-ocr；仅在用户明确同意时再用本地回退方案。
.claude/skills/anbeime-law-to-markdown/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-07 | ✗→✓ | ▲ Improved | -15% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -14% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 38% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -13% | 0% |
| case-12 | ✗→✓ | ▲ Improved | -3% | 0% |
.txt：直接转存为 .md。.pdf / .docx：mineru-ocr skill。mineru-ocr，安装地址：https://github.com/cat-xierluo/legal-skills/tree/main/skills/mineru-ocr
mineru-ocr 处理。mineru-ocr 配置/Token。python-docx / pdfplumber）。--law-decision auto）。###############（仅“第X条”为标题；第X条【条标】整行为标题不拆）Stage2: rejected (non-law-document)。stage2 相对 stage1 的文字内容准确性（去标题符号与空白后字符流必须一致）--law-decision law：调用方已判定为法律文本，直接按法律结构优化。--law-decision non-law：调用方已判定为非法律文本，阶段二直接拒绝。--law-decision auto：不传判定时的默认模式，使用脚本内硬规则。--skip-stage3-check：跳过第三阶段检查（默认不跳过）。--stage3-max-retries：失败后自动重走次数，默认 2。--stage3-strict / --no-stage3-strict：--artifact-level minimal|standard|debug：minimal（默认）：面向交付，输出最少standard：保留过程文件（stage1/stage2/stage3-check）便于排查debug：保留全部过程产物（包含调试信息）markdown/ 子目录，并按输入文件名创建独立目录：markdown/<文件名>/--artifact-level minimal）输出：<原文件名>+审核报告.md（详细过程和结论）<原文件名>+最终成果.md（仅法律文本且审核通过时生成）<原文件名>+审核报告.md<原文件名>+最终成果.md<文件名>.stage1.md<文件名>.stage2.md<文件名>.stage3-check.md--artifact-level standard：保留 stage1/stage2/stage3-check--artifact-level debug：保留全部调试产物--out 或 --out-dir。bashpython3 law-to-markdown/scripts/law_to_markdown.py "input.txt" python3 law-to-markdown/scripts/law_to_markdown.py "input.docx" python3 law-to-markdown/scripts/law_to_markdown.py "input.pdf"
用户明确同意回退时：
bashpython3 law-to-markdown/scripts/law_to_markdown.py "input.docx" --allow-fallback python3 law-to-markdown/scripts/law_to_markdown.py "input.pdf" --allow-fallback
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 11,303 | 7,383 | -35% | 1 | 1 | 0% | 1,860 | 1,554 | -16% | 0 | 0 | — |
case-02 | fail→fail | 15,431 | 6,294 | -59% | 1 | 1 | 0% | 2,770 | 1,575 | -43% | 0 | 0 | — |
case-22 | pass→pass | 9,024 | 2,491 | -72% | 1 | 1 | 0% | 1,561 | 1,602 | +3% | 0 | 0 | — |
case-03 | fail→fail | 14,666 | 3,237 | -78% | 1 | 1 | 0% | 3,178 | 1,650 | -48% | 0 | 0 | — |
case-04 | pass→fail | 14,128 | 7,855 | -44% | 1 | 1 | 0% | 2,299 | 2,427 | +6% | 0 | 0 | — |
case-05 | pass→pass | 22,071 | 17,370 | -21% | 1 | 1 | 0% | 3,677 | 4,121 | +12% | 0 | 0 | — |
case-06 | fail→fail | 12,626 | 10,197 | -19% | 1 | 1 | 0% | 2,263 | 2,834 | +25% | 0 | 0 | — |
case-07 | fail→pass | 15,236 | 5,700 | -63% | 1 | 1 | 0% | 2,623 | 2,218 | -15% | 0 | 0 | — |
case-08 | fail→pass | 14,663 | 4,357 | -70% | 1 | 1 | 0% | 2,323 | 1,993 | -14% | 0 | 0 | — |
case-09 | fail→pass | 6,926 | 3,899 | -44% | 1 | 1 | 0% | 1,377 | 1,896 | +38% | 0 | 0 | — |
case-10 | fail→pass | 12,721 | 4,152 | -67% | 1 | 1 | 0% | 2,240 | 1,943 | -13% | 0 | 0 | — |
case-11 | pass→pass | 14,860 | 7,234 | -51% | 1 | 1 | 0% | 2,413 | 2,509 | +4% | 0 | 0 | — |
case-12 | fail→pass | 11,028 | 3,571 | -68% | 1 | 1 | 0% | 1,977 | 1,925 | -3% | 0 | 0 | — |
case-13 | fail→pass | 10,642 | 2,778 | -74% | 1 | 1 | 0% | 1,880 | 1,697 | -10% | 0 | 0 | — |
case-14 | fail→pass | 8,687 | 2,770 | -68% | 1 | 1 | 0% | 1,442 | 1,682 | +17% | 0 | 0 | — |
case-15 | fail→pass | 9,255 | 2,408 | -74% | 1 | 1 | 0% | 1,626 | 1,583 | -3% | 0 | 0 | — |
case-16 | pass→pass | 8,741 | 1,773 | -80% | 1 | 1 | 0% | 1,515 | 1,481 | -2% | 0 | 0 | — |
case-17 | fail→pass | 10,943 | 3,284 | -70% | 1 | 1 | 0% | 1,960 | 1,715 | -13% | 0 | 0 | — |
case-18 | fail→pass | 9,946 | 2,997 | -70% | 1 | 1 | 0% | 1,758 | 1,714 | -3% | 0 | 0 | — |
case-19 | fail→pass | 14,997 | 3,356 | -78% | 1 | 1 | 0% | 2,421 | 1,783 | -26% | 0 | 0 | — |
case-20 | fail→pass | 13,387 | 6,373 | -52% | 1 | 1 | 0% | 2,047 | 2,355 | +15% | 0 | 0 | — |
case-21 | fail→pass | 15,188 | 4,698 | -69% | 1 | 1 | 0% | 2,372 | 2,096 | -12% | 0 | 0 | — |
case-23 | fail→pass | 12,997 | 6,291 | -52% | 1 | 1 | 0% | 2,234 | 2,303 | +3% | 0 | 0 | — |
case-24 | fail→pass | 10,744 | 3,567 | -67% | 1 | 1 | 0% | 1,917 | 1,754 | -9% | 0 | 0 | — |
case-25 | pass→pass | 7,579 | 2,320 | -69% | 1 | 1 | 0% | 1,298 | 1,559 | +20% | 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. 25 cases were attempted, and 23 counted toward the lift figure. The other 2 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 +56 percentage points is the difference between those two pass rates over the 23 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.