Review Chinese Legal Docs with the legal-review Agent Skill

Where this fits

A developer is building an agent that helps Chinese-speaking users review contracts and legal materials. The user uploads a lease agreement and asks "What clauses could hurt me in court?" Without a dedicated skill, the agent must guess how to structure extraction and risk scoring. The legal-review skill provides a predictable, offline workflow for parsing Chinese legal text, flagging high-risk clauses, and outputting structured review notes—enough for routing validation before you connect a real legal API.

Why agents benefit

• Detects Chinese legal document sections (contracts, complaints, judgments) and classifies their type before deeper analysis.
• Extracts key entities: parties, amounts, deadlines, jurisdiction, and liability clauses.
• Applies rule-based heuristics to flag common litigation risks (e.g., unbalanced indemnity, vague dispute resolution).
• Outputs a structured review summary that can be passed to a larger LLM pipeline or a human reviewer.
• Runs fully offline as a synthetic fixture, so you can test agent logic without real case data or network calls.

Practical scenarios

Contract clause risk pre-screening

Before sending a contract to a lawyer, your agent runs legal-review to identify sections that often hide unfavorable terms. The skill returns a checklist of risky clause types, letting you prioritize which paragraphs need human attention.

Litigation document triage

A user uploads a court notice or complaint. The agent uses legal-review to extract the case number, parties, claims, and key dates, then routes the case to the appropriate internal workflow—tracking deadlines, drafting a response, or requesting more documents.

Parallel document comparison

You have two versions of a lease. legal-review normalizes each into the same structured schema, so your agent can diff clauses and highlight changes in liability, rent adjustment, and termination conditions.

Add it to your agent workflow

Install the skill with npx:

npx skills add Sven-Mirana/sublation --skill legal-review

Then use it in your agent tool definition:

from skills import legal_review

result = legal_review.run(
    document="...",           # 简体中文法律材料
    mode="contract_review"    # 或 "litigation_risk"
)
print(result.risk_flags)

Compare before adopting

Before committing, compare legal-review with alternatives on four axes: quality of risk rules (does it cover the contract types you handle?), maintenance freshness (are issues and PRs being addressed?), replacement options (a general LLM with a prompt may do the same for small volumes), and workflow fit (does its output schema match your downstream tools?). The skill is a synthetic fixture—do not rely on it for real legal advice.

Why it is worth tracking

With 277 stars and a focused scope, legal-review is a good signal that the agent-skill ecosystem is maturing around practical, non-English legal tasks. It is worth tracking because it demonstrates a pattern: lightweight, domain-specific review skills can be composed into larger agent workflows. Evaluate it when your agent needs to handle Chinese legal input but you are not ready to integrate a commercial legal API. See more context on OpenAgentSkill.