Mainland China Court Case Retrieval Report Plus

Problem

In legal research involving mainland China courts, case retrieval is often blocked by concrete issues: whether results cover the right jurisdiction, date range, and dispute focus; how to compare outcomes quickly among many candidates; how to surface authoritative cases; and how to filter cases by a specific lawyer or firm when the API lacks direct filter fields.

How It Works

The skill calls the Yuandian legal data MCP, using yuandian_case_vector_search for case retrieval and yuandian_rh_case_details when full details are needed. Its workflow includes:
- Parameter collection: collects jurisdiction, date range, keywords/dispute focus, and representation scope via AskUserQuestion, then echoes the conditions before retrieval.
- Filtering: supports exact province via xzqh_p, specific court via fayuan, cause of action via ay, and date bounds via ja_start/ja_end. When no jurisdiction is selected, it runs a single nationwide retrieval and locally groups results by court or fayuan.
- Retrieval optimization: sets rewrite_flag: false for structured queries to skip query rewriting; returns a larger first batch by mode, such as return_num = 15 for nationwide mode, 10 for precise province/city mode, and 15 for lawyer/firm scope, reducing back-and-forth turns.
- Review and report: renders an interactive widget, prefetches authoritative case details, extracts judgment outcome summaries, and marks public-report, guiding, typical, or excellent cases. The user chooses whether to continue or generate the report; all batches are merged and deduplicated by case number.
- Representation scope: for a specific lawyer plus law firm, it combines query bias with post-retrieval content filtering, requiring the lawyer's name and the firm's full name or derived short form to appear in the same agency context. Lawyer-only or firm-only modes use looser matching.

Limits and Notes

• The Yuandian legal data MCP must already be connected.
• Court documents may anonymize counsel, so searches by a specific lawyer or firm may have low recall; if filtering yields no cases, the skill should explain why rather than produce an empty report.
• It is useful for dispute-focus research, similar-case comparison, and initial screening, but it does not replace manual verification of specific case numbers.