医生工作台助手 (Agentic AI 科研平台)

• Outcome: Pass
• Latest audit: May 29, 2026 at 8:31 PM UTC
• Version:
  0.1.0

Overview

This skill is a narrow doctor-workbench helper that reads one local demo patient file and posts generic progress updates to a local service.

Install this only where the localhost reporting endpoint is expected and the referenced demo patient JSON is authorized for use. Treat any patient-related output as sensitive, verify that no real patient identifiers are exposed unintentionally, and have a clinician review generated follow-up tasks before acting on them.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access
• Supply Chain: Unpinned Dependencies, External Script Fetching, Obfuscated Code
• Excessive Agency: Unrestricted Tool Access, Autonomous Decision Making, Scope Creep

Findings (1)

Missing User Warnings

Medium

• Confidence: 95% confidence
• Finding: The skill explicitly posts workflow status to a local HTTP endpoint without informing the user or documenting what data may be transmitted. Even though the payload shown is status metadata rather than patient content, this is a healthcare-context skill and silent network reporting around patient-processing activity creates privacy, audit, and trust risks, especially if the local service logs, forwards, or correlates requests with patient sessions.

VirusTotal

61/61 vendors flagged this skill as clean.