moot-court-ai

What This Skill Does

Moot Court AI is a specialized OpenClaw skill designed to simulate a realistic Chinese civil court proceeding. By utilizing a multi-agent architecture controlled by a deterministic Lobster workflow, the skill orchestrates four distinct roles: Clerk, Plaintiff, Defendant, and Judge. The system strictly adheres to the legal phases mandated by the Civil Procedure Law of the PRC, ensuring that every session progresses logically from the opening phase to the final judgment. It leverages high-performance models like DeepSeek and Qwen to ensure legal reasoning is both coherent and contextually accurate, making it an essential tool for legal professionals, law students, and researchers looking to stress-test legal arguments or analyze litigation strategies in a controlled virtual environment.

Installation

To begin using the Moot Court AI, ensure you have the OpenClaw environment properly initialized. Install the skill directly from the repository using the command: clawhub install openclaw/skills/skills/baobaodawang-creater/moot-court-ai. Before running, you must configure the following environment variables in your workspace: DEEPSEEK_API_KEY and DASHSCOPE_API_KEY. Once the keys are set, verify your connection to the model providers, as the simulation requires active API calls to maintain the dialogue flow across the four agents.

Use Cases

• Legal Training: Law students can use this tool to practice responding to 'three-validity' (三性) challenges and improving oral advocacy skills.
• Case Strategy Simulation: Litigation lawyers can input actual case files to evaluate the strengths and weaknesses of their claims and defenses before entering a real courtroom.
• Academic Research: Researchers can analyze how different LLMs interpret specific clauses of Chinese civil law, facilitating comparative studies on legal AI reasoning.
• Mock Trials: Organizations can host internal moot court sessions to resolve disputes or train junior staff in civil litigation procedures.

Example Prompts

1. "Initialize the moot court session for the contract dispute case located in /workspace/case-files/contract-001."
2. "Run the court simulation using Qwen-Max for the Judge and DeepSeek-Reasoning for the counsel agents, then export the transcript to PDF."
3. "Summarize the primary points of contention raised by the defendant during the evidence exchange phase of our current simulation."

Tips & Limitations

To maximize the quality of the output, ensure that your input documents, such as complaint.md and defense.md, are structured clearly with bullet points and explicit legal claims. The simulation is deterministic in its flow; do not interrupt the Lobster workflow while the agents are in the 'evidence exchange' phase to avoid state synchronization errors. Note that while the model is highly capable, it is a simulation tool and not a replacement for professional legal advice. Always review the final judgment rendered by the AI against existing statutes and local judicial interpretations.