
from china-lawyer-analyst128
Automated Chinese-law analysis skill that performs three-phase legal analysis (initial analysis, legal verification, and reflection/correction) with dynamic jud
This skill provides structured legal analysis from the perspective of Chinese law. It automates a three-phase workflow: Phase 1 — initial analysis (IRAC + 10-step method and case-type identification), Phase 2 — automated legal verification (checks statutes, judicial interpretations, and versioning), and Phase 3 — reflection and correction (updates citations and re-runs analysis). It includes a dynamic judicial-interpretation index to load only the necessary statutory commentary, reducing token and compute costs while preserving accuracy. Outputs include a structured opinion, a six-part requirements checklist, and a verification report.
Use when you need rapid, citation-aware legal analysis for Chinese law issues such as contract disputes, tort claims, company law questions, construction disputes, investment/financing issues, labor law, IP, or litigation/ arbitration scenarios. Ideal when accuracy against law versions matters (e.g., check whether a statute was amended) or when you want a reproducible, checklist-driven legal opinion.
Designed for agent platforms that support long-form procedural skills and tooling integrations (Claude Code / Claude Agent SDK, similar LLM agent runtimes).
China law analysis skill with a sophisticated MoE + dynamic index architecture for token-efficient legal reasoning. Three scripts bundled: two fail at import because they depend on db_accessor from the tools/ directory which isn't included in the scripts bundle; one (init_case_db.py) runs fine. SKILL.md is extremely detailed (~3800 lines of content) covering 8 legal domains, 45 case types, and a three-phase automated verification workflow.
db_accessor (from tools/ directory, not included in scripts bundle)Well-architected skill with thoughtful token optimization. The scripts dependency issue (tools/db_accessor.py not in scripts/) is a packaging gap rather than a code defect. No security concerns — all local SQLite operations, no network calls, no credentials.