SZZZ Case Study Lite｜案例分析技能

• Outcome: Pass
• Latest audit: Jul 23, 2026 · 5:08 AM UTC
• Version: 3.0.1

Overview

The skill locally processes user-selected legal documents and writes a project case library, with cautions around generated sensitive copies, unlocked dependencies, and a reset command for the results folder.

Install only if you are comfortable with the skill reading every supported legal document in the folder you provide and creating raw text copies and reports under law_analysis_results. Avoid pointing it at overly broad or unrelated folders, review generated outputs for sensitive material, and use reset carefully because it deletes the generated analysis results directory.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Supply Chain: Unpinned Dependencies, External Script Fetching, Obfuscated Code
• MCP Least Privilege: Underdeclared Capability, Wildcard Permission, Missing Permission Declaration
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access

Findings (6)

Lp3

Medium

• Category: MCP Least Privilege
• Confidence: 91% confidence
• Finding: The skill clearly instructs execution of local Python scripts and relies on file reads, file writes, and shell execution, yet no permissions are declared. This creates a transparency and consent gap: a user or host system may not realize the skill will modify local files and invoke local tooling, increasing the risk of unintended filesystem changes or unsafe script execution.

Missing User Warnings

Medium

• Confidence: 95% confidence
• Finding: The skill directs the agent to run scripts that create and update analysis outputs, reports, and derived files in the local project tree, but it does not explicitly warn the user about these filesystem changes before execution. In a document-processing workflow, this can lead to silent writes, accidental overwrites, or creation of sensitive derived artifacts without informed consent.

Missing User Warnings

Low

• Confidence: 90% confidence
• Finding: The prompt explicitly instructs the agent to invoke a file-writing tool and save output into the local workspace without any user-confirmation guard. In this skill, writing a report file is part of the intended workflow, so this is not overtly malicious, but it still creates an integrity risk because the agent is being directed to modify local files automatically based only on prompt content.

Missing User Warnings

Medium

• Confidence: 89% confidence
• Finding: The reset command recursively deletes the entire analysis-results directory with shutil.rmtree and performs no confirmation, dry-run, or path safety checks beyond computing the expected subdirectory. In a skill/agent context, accidental invocation or misuse can irreversibly destroy prior work product and audit artifacts, which is especially risky because this tool manages user-generated case analysis data.

Unpinned Dependencies

Low

• Category: Supply Chain

Content

# 安装方法：python3 -m pip install -r requirements.txt

# PDF 解析
pdfplumber>=0.10.0

# DOCX 文档读取
python-docx>=0.8.11

• Confidence: 87% confidence
• Finding: pdfplumber>=0.10.0

Unpinned Dependencies

Low

• Category: Supply Chain

Content

pdfplumber>=0.10.0

# DOCX 文档读取
python-docx>=0.8.11

# 旧式 DOC：macOS 使用系统 textutil；其他系统需另行安装 LibreOffice。

• Confidence: 86% confidence
• Finding: python-docx>=0.8.11

VirusTotal

39/39 vendors flagged this skill as clean.

Static analysis

No suspicious patterns detected.