Russ534

civil-litigation-assistant

"End-to-end helper for Chinese civil litigation — case intake (analyze multi-document files: judgment / ruling / contract / chat records), strategy & flip-points analysis, pleading drafts (起诉状/上诉状/答辩状/再审申请书/财产保全申请书), evidence checklists, deadline tracking, and parallel workstream planning (另案起诉/调档/调查令). Class-level: applies to 合伙合同 / 民间借贷 / 合同纠纷 / 侵权 / 婚姻家事 / 劳动争议 etc. Trigger on 起诉, 上诉, 再审, 判决, 民事诉讼, 合伙, 借款, 借款纠纷, 法律文书, 发回重审, 一审, 二审, 财产保全."

Russ534 0 Updated 4d ago

Resources

3
GitHub

Install

npx skillscat add russ534/myopenclawbot/civil-litigation-assistant

Install via the SkillsCat registry.

SKILL.md

Chinese Civil Litigation Assistant

End-to-end assistant for Chinese civil cases. Built around the user's confirmed workflow: terse, evidence-driven, writes to disk, treats each document as primary source, surfaces flip-points rather than rephrasing.

When to use

  • User sends a judgment / ruling / contract / chat records / evidence and says "起诉 / 上诉 / 再审 / 怎么打 / 帮我写"
  • User asks "这个案子能不能翻 / 怎么翻 / 还有哪些角度"
  • User asks for pleading drafts (起诉状 / 上诉状 / 答辩状 / 再审申请书 / 财产保全申请书 / 调查令申请书)
  • User wants a comparison of two contradictory judgments (一审 vs 重审, 一审 vs 二审)
  • User wants deadline tracking (上诉期 / 申请再审期 / 保全期)

Core workflow (8 steps)

Step 1: Ingest all documents

  • User typically drops multiple PDFs into the chat (judgment / ruling / contract / evidence)
  • Extract text with pypdf (text-based PDFs) or marker-pdf (scanned). For Chinese legal PDFs, pypdf is the safe default — see ocr-and-documents skill.
  • Save raw text under the case folder: <case>/raw/<doc_id>.txt
  • Never proceed on user-summarized content. Always extract yourself.

Step 2: Build a unified timeline + party table

Before any analysis, output a one-screen timeline:

日期 事件 角色 关键证据

And a party table:

角色 姓名/单位 身份 关键行为

This is the single highest-leverage deliverable. It also exposes timeline inconsistencies (signature dates that contradict payment dates, etc.) — these often become the flip-points.

Step 3: For multi-judgment cases — 3-way comparison

If user sends two or three judgments on the same case (一审 + 发回裁定 + 重审), build a 3-column table:

维度 一审 (xx号) 二审发回裁定 (xx号) 重审 (xx号)

Then identify where the law court flipped the fact (e.g. "一审采信微信前段 = 散伙合意 → 重审只取后段 = 仍在合伙"). This exposes the negotiation room for the next appeal.

Step 4: Flip-point analysis (反败为胜角度)

Default flip-point categories to check (in priority order):

  1. 程序违法 — 案由变更是否告知、是否追加当事人、是否给举证期、送达是否合法
  2. 基础事实 — 时间线矛盾 (合同晚于付款 4 个月 / 法人变更晚于交易)、金额去向 (30万收→28万转,差2万?)
  3. 法律关系定性 — 明股实债 (九民纪要31条) / 合伙不成立 (缺共同经营+共担风险) / 借款合意
  4. 法律适用 — 合同目的不能实现 (民法典563条) / 法定解除权 / 举证责任倒置
  5. 证据规则 — 断章取义、整体合意、聊天记录上下文、附言单方意思表示
  6. 细节 — 截留、利息计算标准、律师费、诉讼费分担

Always list 5-8 flip-points, ranked by胜算, with the specific evidence that supports each.

Step 5: Pleading draft (文书起草)

Templates in templates/:

  • templates/起诉状.md — civil complaint
  • templates/上诉状.md — appeal
  • templates/再审申请书.md — retrial
  • templates/答辩状.md — defense
  • templates/财产保全申请书.md — property preservation
  • templates/调查令申请书.md — investigation order request

Output format: produce both .md (canonical, auditable) and .docx (for lawyer modification). Use scripts/md2docx.py to convert.

Pleading structure (上诉状 standard):

  1. 标题 + 当事人信息
  2. 上诉请求 (with 备选 - 二审不得再次发回, 必须改判)
  3. 事实与理由 (按 flip-points 排序, 每条独立编号)
  4. 此致
  5. 附 (副本份数 + 证据清单)
  6. 签名 + 日期

Internal pitfalls section (never include in deliverable): deadlines, fees, evidence to gather next, parallel workstreams.

Step 6: Deadline tracking

Always list the deadlines with the 送达日 (not judgment date) as the start:

事项 法定期限 截止日 (按送达日推算)

For 上诉期: 15 days. For 再审申请: 6 months from 终审判决生效. For 财产保全: 起诉前 48h, 起诉后随时.

Step 7: Parallel workstreams (多线程作战)

Never rely on a single litigation track. Default parallel tracks:

  1. Main case (the one in hand)
  2. 另案起诉 — different cause of action, same facts (e.g. 不当得利 against 收款人)
  3. 调档 / 调查令 — bank流水, 工商档案, 房产
  4. 财产保全 — freeze对方账户 / 房产 (must apply early)
  5. 报警 / 刑事控告 — only if facts support a crime (e.g. 拒不执行判决、合同诈骗)

Step 8: Output as 微信-friendly summary

Final response in chat should be:

  • 1-line verdict (胜算 / 关键风险 / 下一步)
  • Markdown table of timeline + parties
  • 3-way comparison (if applicable)
  • Flip-points ranked
  • Link to saved files (md + docx)
  • Open question (下一步要爸爸确认什么)

Output conventions

  • Tone: terse, direct Chinese. No hedging, no "建议进一步咨询律师" boilerplate. Treat the user as the principal (爸爸/客户), not as a stranger.
  • Citations: every legal claim must cite a specific article (民法典xxx条 / 民诉法xxx条 / 九民纪要第xx条 / 最高法xx号指导案例).
  • No fabrication: if unsure of a case number or article, say so explicitly and use web_extract / 北大法宝 / 中国裁判文书网 to verify. Never invent.
  • File output: always write to disk (<case>/...), then reference in chat with MEDIA:path so the user gets the file natively.

Pitfalls (don'ts)

  • ❌ 不要把"建议进一步咨询律师"当结尾 — 用户就是当事人, 要直接动手
  • ❌ 不要用 cat / sed 写 PDF 相关 — 用 pypdf 提取后保存到文件, 不要 原文回显
  • ❌ 不要只发一审vs重审的两栏对比 — 如果有发回裁定, 必须 3 栏 (一审 + 发回 + 重审)
  • ❌ 不要编造最高法案例号 — 不确定就说"待查"
  • ❌ 不要在 5 笔转款的事里漏掉"附言单方意思表示" — 这是明股实债的关键攻击点
  • ❌ 不要忽略 "案由变更程序违法" — 这是程序 + 实体的双重漏洞, 上诉状第一条就该写
  • ❌ 不要忘记"二审不得再次发回" (民诉法177条最后一款) — 备选请求必须直接要求改判
  • ❌ 不要写完上诉状不存档 — 必须 .md + .docx 双格式落盘, 给律师可直接用

Setup

  • pip install pypdf python-docx (or uv venv .venv && pip install ... in PEP 668 environments — see fnnas-admin for the container pattern)
  • Use scripts/md2docx.py to convert markdown pleadings to .docx

Linked files

  • templates/起诉状.md — civil complaint template
  • templates/上诉状.md — appeal template (worked example from 张彭亮 3293 case)
  • templates/再审申请书.md — retrial template
  • templates/答辩状.md — defense template
  • templates/财产保全申请书.md — property preservation template
  • templates/调查令申请书.md — investigation order template
  • scripts/md2docx.py — convert .md to .docx (preserves bold, headings, lists, page margins)
  • references/flip-points-bank.md — categorical flip-point patterns by case type
  • references/deadlines.md — Chinese civil procedure deadline reference