explaingit

neu-zha/legal-ai-skills

21PythonAudience · writerComplexity · 3/5Setup · moderate

TLDR

A set of AI agent skill files for Chinese legal work, preventing fabricated citations, fact-checking drafts, formatting Word documents, and querying the PKULaw database.

Mindmap

mindmap
  root((legal-ai-skills))
    Core Problem Solved
      Prevents fake citations
      Marks unverifiable claims
      No invented case names
    Skills Included
      Fact-checking draft
      Citation formatting
      Word doc formatting
      Evidence catalog
      PKULaw database query
    Outputs
      Verified legal drafts
      Formatted footnotes
      Structured Word files
    Audience
      Law students
      Legal researchers
    Agent Runtimes
      Claude Code
      Codex
Click or tap to explore — scroll the page freely

Code map

Detail Auto

An interactive map of this repo's files and how they connect — its source is parsed live in your browser. Click Visualize to build it.

filefunction / class

Things people build with this

USE CASE 1

Have an AI agent scan a Chinese legal analysis draft and flag every unverified citation with a [pending: source] marker instead of guessing.

USE CASE 2

Reformat all footnotes in a law school assignment to match the Chinese Law Journal Citation Handbook second edition standard automatically.

USE CASE 3

Search the PKULaw database for statutes and case law through an AI agent using your institutional access credentials.

Tech stack

Python

Getting it running

Difficulty · moderate Time to first run · 30min

PKULaw database skills require institutional access credentials for the API integration.

In plain English

legal-ai-skills is a collection of reusable instruction sets (called skills) for AI agents working with Chinese legal documents. These skills are designed for agent runtimes that support a SKILL.md-based loading system, such as Claude Code, Codex, or similar tools. You install them by copying skill folders into your agent's skills directory, then ask the AI to use a specific skill when you give it a task. The core problem the project addresses is that AI tools often fabricate legal citations: inventing article numbers, case names, or page references that do not exist. The skills in this collection are structured to prevent that. When a piece of information cannot be verified, the agent is instructed to mark it with a placeholder like [pending: source] rather than guess. This applies to statute references, case citations, and legal footnotes. The skills cover several workflows. A fact-checking skill examines legal analysis drafts against source materials the user provides. Citation skills check and reformat footnotes following the Chinese Law Journal Citation Handbook (second edition). A formatting skill produces properly structured Word documents for law school assignments and legal writing. An evidence catalog skill organizes case materials into a structured list. Several skills connect to PKULaw, a major Chinese legal database, to retrieve statutes, regulations, and case law directly through an API integration, though users need their own institutional access credentials to use those. Word document editing is also covered, with skills for handling footnotes, tables of authorities, and document structure in .docx files. The README is detailed and includes quick-start instructions aimed at law students with no prior technical setup experience.

Copy-paste prompts

Prompt 1
Load the legal-ai-skills fact-checking skill and scan this Chinese contract law analysis for any article citations that can't be verified against the Civil Code source I'm providing.
Prompt 2
Using the citation skill from legal-ai-skills, reformat all the footnotes in my attached document to follow the Chinese Law Journal Citation Handbook second edition format.
Prompt 3
Use the PKULaw skill from legal-ai-skills to search for all judicial interpretations related to Article 1165 of the Chinese Civil Code published after January 2021.
Open on GitHub → Explain another repo

← neu-zha on gitmyhub — every repo by this author, as a profile.

Verify against the repo before relying on details.