legal-case-analysis

legal-case-analysis

General legal analysis Skill. Used to organize scattered case materials into verifiable, retrievable, arguable, and reusable legal analysis results.

This skill is first and foremost a legal analysis engine, and only secondarily generates reports, consultation replies, client explanations, or retrieval task lists as needed. Legal analysis can be used as a preliminary step for other Skills.

This skill is not bound to a specific cause of action. Specific domain rules are dynamically supplemented through case materials, legal retrieval results, and user objectives.

Core Principles

Organize the facts first, then apply the law.
Distinguish evidence from statements first, then form judgments.
First use yuandian-law-search to complete basic legal retrieval, then conduct legal analysis based on the retrieval results.
Raise retrieval questions first, then cite legal bases.
For litigation scenarios, prioritize element-based analysis: basis of claim, constitutive elements, facts, evidence, burden of proof, and defenses corresponding item by item.
Every important conclusion should correspond to a factual source, evidentiary source, or retrieval basis.
For information not mentioned in the materials, mark it as "not mentioned" or "to be supplemented"; do not fill it in on your own.
For disputed materials, evaluate authenticity, relevance, legality, and purpose of proof at the same time.
For issues involving limitation periods, periods, old and new laws, or changes in judicial interpretations, they must be verified separately.
Do not write specific legal article numbers, full names of judicial interpretations, or case numbers from memory. Article numbers must come from the retrieval results of this session or materials provided by the user; when no retrieval has been conducted, write only the content of the rule and the name of the law (e.g., "the provisions of the Copyright Law on statutory damages"), and mark "article number pending retrieval". This applies to the entire report, not limited to the jurisdiction section—article numbers from memory look the most professional, but they are the most dangerous errors.

Input

One or more of the following inputs may be accepted:

Case overview, consultation records, client statements.
Contracts, letters, payment vouchers, registration materials, judgment documents, enforcement materials.
Evidence lists, chat records, audio transcriptions, meeting minutes.
Legal provision retrieval results, case retrieval results, legal research reports.
Analysis objectives specified by the user, such as "assess whether to sue", "generate a legal analysis report", "form a client explanation".

If the materials are PDFs, images, or scanned copies, first use MinerU OCR, PaddleOCR, or other document conversion tools to convert them to Markdown before analysis.

Output

The default output is Markdown. Whether to generate a formal report depends on the user's objective; many scenarios only require analysis drafts, consultation replies, or analysis conclusions for use by other Skills.

Desensitized Output Mode (Optional by Scenario)

By default, mandatory desensitization is not performed. This Skill is oriented toward real case handling: materials are meant to be submitted to courts, arbitration, or clients. Using real parties and place names as-is is a prerequisite for analytical accuracy; forced desensitization would instead distort them and lose the geographic information needed for connection points and jurisdiction determinations. ThereforeUnless the user explicitly requests it, output directly using real information。

Only when the user states it is forcourses, competitions, public demonstrations, case sharing, internal training, or public disclosureshould redacted output mode be enabled (first read references/anonymization-output.md before writing):

Anonymize party names, company names, contact information, addresses, residence/business locations, administrative divisions, and other fields that can directly identify or, in combination, identify a subject.
Abstract the region to "a certain city" or "Location A"; granularity at the district/county level and below (including "a certain district," "a certain county," "a certain street") must not appear at all.
Preserve the legal relationships, factual structure, and points of dispute—these are where the analytical value lies and must not be lost due to redaction.
Redaction does not change the core principles of "marking information not mentioned in the materials as not mentioned/to be supplemented" and "distinguishing facts from statements" (see core principles 2, 6, and 7).

When the materials themselves are already fictional or already redacted, do not perform secondary anonymization. Competition questions, textbook cases, and already public teaching judgments often themselves use fictional subject names (such as "Lin Xiaotang" and "Hengrui Real Estate Co., Ltd.") and have already abstracted the region to "a certain city" and masked identity information as "3207XXXX...". In such cases, you shouldfollow the existing naming and granularity of the materials as they are, and must not rewrite "Lin Xiaotang" as "A," "certain A," or the like—secondary anonymization adds no protection and only destroys the continuity of the subject chain, making it difficult to review the correspondence of parties, third-party relationships, and priority of rights. Judgment criterion: if a name in the materials is already a fictional code name or has been explicitly declared fictional, it is deemed to have satisfied the redaction requirement. The security baseline (irreversible sensitive fields) still applies separately: masked or semi-masked ID numbers and mobile phone numbers are retained in their existing form in the materials, and are not completed or expanded.

Security baseline (applicable in any scenario, regardless of redaction mode)

The followingirreversible sensitive fieldsshould not be output as-is to externally visible content regardless of whether redaction mode is enabled: ID numbers, bank account numbers, private personal contact information (private mobile phone numbers/private email addresses), and payment password-type credentials. If the materials contain such fields and the output is for external use, mask them or leave them blank and mark "omitted"; for purely internal case work drafts, the user may decide whether to retain them. This baseline is not equivalent to "redacting the entire case"; it targets only a very small number of fields that are irreversible once leaked.

Choose the output according to the usage scenario:

Analysis draft: for subsequent writing, plan generation, litigation preparation, or internal discussion, emphasizing facts, elements, evidence, research, and risks.
Brief consultation reply: for everyday legal consultations with fewer facts and a single objective, emphasizing the conclusion, main reasons, risks, and next steps.
Internal analysis report: for the lawyer team, emphasizing facts, evidence, legal reasoning, risks, and strategy.
Client explanation version: for non-legally-trained clients, expressed more plainly, emphasizing conclusions and action recommendations.
Research task list: for yuandian-law-search or zhihe-legal-research。

Work method

Place the analysis process and work method in references/ , and read as needed for the task:

references/iteration-context.md: long-term iterative context, including positioning, boundaries, resource layering, and subsequent evolution direction.
references/analysis-engine-boundary.md: the positioning of the legal analysis engine, as well as its collaboration boundaries with legal-proposal-generator and other Skills.
references/workflow.md: the complete workflow, including objective identification, case stage, material handling, research, argumentation, and delivery check.
references/report-depth-control.md: Select simplified consultation, standard analysis, or phased report for complex cases according to case complexity.
references/element-litigation-nine-step.md: The nine-step element-based litigation analysis method, used forcivil/commercialanalytical framework for litigation and dispute resolution scenarios.
references/criminal-case-analysis.md: Criminal case analysis method (the two-tier structure of unlawfulness and culpability as the backbone, prosecution review skeleton + defense adversarial testing), used for criminal case assessment, working drafts for prosecution review/arrest review, characterization and sentencing analysis; objective unlawfulness first, then subjective culpability, progressing layer by layer, with exoneration if any layer is not satisfied.
references/fact-evidence-extraction.md: Rules for fact and evidence extraction.
references/civil-commercial-analysis-workflow.md: Steps for legal analysis of complex civil and commercial cases.
references/issue-spotting.md: Rules for legal issue identification and issue tree decomposition, including the rule for exhausting rights items in rights-bundle-type causes of action (separate listing of personal rights items/property rights items), rules for identifying procedural paths (primary/parallel/subsequent), and linkage with jurisdiction analysis.
references/legal-research-bridge.md: Rules for linkage with legal research.
references/legal-research-result-ingestion.md: Backfill rules when the user has already provided a research archive (status marking, source level, outdated interception).
references/time-effectiveness-check.md: Fixed output table for transition between old and new law and limitation-period checks under multiple time baselines.
references/argument-risk-assessment.md: Rules for argumentation and risk assessment.
references/anonymization-output.md: Triggering of anonymized output mode, field list (including region/residence and combination identification risks), retained items, and boundaries (for courses/competitions/public demonstrations).

Output templates

Final delivery templates are placed in templates/ :

templates/simple-consultation-answer.md: Simplified consultation reply template.
templates/legal-analysis-report.md: General legal analysis report template, including fixed anchor points: Section 6.1 temporal effect check (multiple baselines T1/T2/T3 mandatory), Section 8.3 claims and amounts ledger (according to damages-calculation-table.md; state not applicable if there is no monetary claim), Appendix analysis process workbook (according to analysis-workbook.md; substantive working drafts must be written rather than pointers).
templates/legal-research-query-list.md: Legal research task list template.
templates/damages-calculation-table.md: Template for multi-category amounts ledger and duplicate calculation risk (contract termination and damages scenarios).
templates/analysis-workbook.md: Analysis process workbook (materials list → facts ledger → issue tree → research → backfill → temporal effect → argumentation → amounts → procedural path), for teaching/competition/retrospective demonstration of the process.
templates/client-summary.md: Client explanation template.