The Agent Skills Directory

[PROMPT_INJECTION]: The skill uses standard instructional language to define a legal analysis workflow. It contains no instructions to bypass safety guidelines, override system constraints, or reveal internal configurations. It includes appropriate disclaimers stating that its output is for reference and does not constitute formal legal advice.
[DATA_EXFILTRATION]: No network communication tools (e.g., curl, wget, fetch) or file system operations are used. There are no patterns indicating that processed case data is sent to external or unauthorized domains.
[REMOTE_CODE_EXECUTION]: The skill is entirely composed of markdown-based instructions and templates. It does not include scripts, download external binaries, or call for the installation of third-party packages.
[COMMAND_EXECUTION]: There is no use of shell commands, subprocess spawning, or dynamic code execution (e.g., eval, exec). The instructions are purely analytical and focus on generating text-based reports.
[CREDENTIALS_UNSAFE]: The file does not contain any hardcoded API keys, tokens, passwords, or other sensitive credentials.
[SAFE]: The skill demonstrates high-quality, structured content with specific legal references (e.g., Chinese Procedure Laws) and follows best practices for legal analysis. It includes logic for identifying 'traps' like unverified photocopies or incomplete electronic data, which enhances its safety profile for users.

evidence-evaluation