Around one judgment, answer "what the court found, on what basis, how the reasoning connects to the result, and which points merit further verification." Retain litigation-analysis invocation name; complete one judgment document node at a time, without expanding into full-case analysis.
Input and Scope of Responsibility
The necessary input is one readable judgment. The user's goal, focal issues, represented party, and reader are optional context; when no position is specified, analyze neutrally and do not assume by default that the plaintiff is our side. A judgment can be analyzed independently; there is no need to request the complaint, original evidence, or the entire case file just to begin.
- For scanned copies, first convert them into locatable text using the currently available document reading/OCR tools, and preserve page numbers. Processing the same file does not constitute multi-material analysis.
- When the user provides multiple judgments and asks for separate interpretations, output each independently; when comparison across documents or comprehensive assessment with other evidence is requested, defer to
legal-case-analysisto lead. - When the judgment and attachments are provided together but only interpretation of the judgment is requested, read only the target judgment. If the target judgment is unclear, first determine the target and do not automatically scan other files in the same directory.
- When the input contains only a complaint, evidence, or trial transcript, explain that it should be redirected to comprehensive analysis; if the target Skill is unavailable, provide the applicable entry point and handover information, and do not falsely claim to have invoked it.
- When only the "This Court finds" section or an excerpt of the operative part is available, complete the analysis of the visible parts and mark it as "partial analysis"; missing sections must not be written as undisputed or without omission.
Processing Workflow
1. Establish document positioning and source hierarchy
Record the file, case number, court, cause of action, instance level, judgment date, and readable scope. Locate by page number/paragraph number; when there are no page numbers, locate by section and short quotations, and do not fabricate line numbers.
Distinguish throughout: parties' claims, court findings, evidence recited by the judgment, analyst inferences, and matters not mentioned. A statement in the text that "the plaintiff submitted a contract" does not mean the original contract has been reviewed; a second-instance recital of the first instance does not mean the original first-instance judgment has been read.
Load as needed:
- First-instance judgment: read first-instance extraction rules。
- Second-instance judgment: read second-instance extraction rules。
- Judgments at other instance levels: follow this workflow and analyze according to the original text's role, procedure, and operative part, without applying civil first-instance fields.
- Judgments related to intellectual property and similar matters: read according to the issues domain reading questions. It only helps understand the reasoning in the text and does not provide independent conclusions on current law.
2. Reconstruct the adjudicative reasoning issue by issue
Correspond one by one to the claims/defenses, key facts, evidence stated in the judgment, reasons for acceptance, rules cited by the court, application process, and conclusion. Preserve amounts, calculation bases, temporal conditions, allocation of burden of proof, unadopted views and reasons, and anything that affects the outcome.
If the court did not explicitly summarize the issues, the analyst may summarize and mark them. Explain how the decisive findings affect other issues; do not merely restate conclusions or elevate parties' statements into facts.
3. Conduct bounded in-depth analysis
Check whether the facts and reasons in the text connect, whether the acceptance or rejection of evidence is explained, whether key defenses are responded to, whether the calculations are internally consistent, and whether the reasoning and the operative part are consistent. For each doubtful point, write out the location in the original text, the observation, possible counter-interpretations, the impact, and what still needs to be verified.
- You may point out internal contradictions in the text, reasoning that is not developed, and potential disputes; "the judgment does not state" does not equal "the court did not review."
- Cite statutory provisions, judicial interpretations, or cases only by transcribing what the judgment actually records, and mark them as "cited by the judgment, no external verification performed." If the full text of a provision is not in the document, do not fill it in from memory; the judgment's citation itself does not prove that it is currently effective or correctly applied.
- This node does not proactively search the internet, does not collect case files, and does not give the probability of reversal, a complete decision on whether to appeal/retry, a fee quote, or a case-handling plan. When external legal source verification or a comprehensive remedies assessment is needed, hand the issue off to
legal-case-analysis; a pure search objective can be handed off to the currently available legal research Skill. - The Skill does not execute submission, sending, signing/sealing, or writing case status back. Only when the user requests file delivery, write to the specified output location, preserving the original judgment.
4. Explain the operative part and the notice matters
Explain item by item the operative part, the obligor, the amount, the conditions for performance, litigation costs, and the notice of remedies in the text. Restate the time limits and trigger conditions recorded in the judgment, do not substitute the judgment date for the date of service, and do not treat fixed numbers of days in a template as the time limit in this case. When the trigger facts are missing, mark it as "the specific deadline cannot be calculated on this basis."
5. Deliver according to the objective and stop at this node
By default, deliver one in-depth judgment analysis. When the user only requests a client explanation or research questions, after completing the necessary in-text analysis, directly deliver the corresponding version, and do not force the generation of three documents:
- Client explanation: explain the judgment outcome and reasons in plain language, preserving uncertainty.
- Research questions to be verified: list the doubtful points in the text and the search propositions, and do not pretend that the search has already been completed.
- Visualization: use only when it aids understanding, and do not add information just to make a chart.
In external content, redact irrelevant ID numbers, bank account numbers, and private contact information; for public teaching/demonstration, de-identify according to the user's purpose. For content involving real case remedies or external delivery, retain the reminder for lawyer review.
Hand off to comprehensive assessment
When further handling is needed, attach a "judgment analysis handoff," without creating a separate complete case report:
| Field | Content |
|---|---|
| Source | Judgment document, case number, trial level, method of locating the original text, and readable scope |
| Completed | Summary of the operative part, decisive issues, court findings, and in-text bases |
| Questions to be verified | Doubtful points, possible counter-interpretations, impact, and the case file/search basis needed |
| Not yet verified | Unread attachments, authenticity of original evidence, effectiveness of external legal sources, service/effectiveness status |
| Follow-up objective | The comprehensive judgment and position requested by the user; if unspecified, mark it truthfully |
If downstream is unavailable, this node can still be completed and the above content delivered; do not write "recommend handoff" as "search/assessment completed." When comprehensive assessment already has traceable conclusions, reuse them and do not repeat the analysis of the entire case.
Complete the check
- Important findings, quotations, amounts, and doubtful points can all be located in the original judgment text; missing parts have been marked.
- All main texts and decisive issues have been covered; the parties' claims, the court's findings, and the analytical opinions are separated.
- The source limitation remains for second-instance citations of first-instance findings and for evidence recounted in the judgment; no guarantee is made as to the current validity of legal sources.
- No unsourced win rates, deadlines, fees, or comprehensive litigation decisions have been generated; research questions have not been passed off as search results.
- Deliver the version that meets the user's requirements; when what is read is an excerpt, do not claim that analysis of the entire judgment has been completed.
Dependencies
Readable-text analysis requires no scripts or third-party packages. Conversion of scanned judgments depends on the currently available reading/OCR tools; when the tools are unavailable, state the gap. Independent judgment analysis does not depend on the installation of other Skills; comprehensive assessment and retrieval are follow-up on-demand entry points.
Maintenance Notes
When adding new domain reading questions, follow domain expansion boundaries, and do not stuff new comprehensive business processes back into this node.