ClawHub
Back to skill

Security audit

unisound-overall-report

Security checks across malware telemetry and agentic risk

Overview

The skill does what it claims, but it handles sensitive medical reports in a way that is not fully matched by its privacy claims.

Review this skill before installing in any real health-data workflow. Use it only with an approved medical-model endpoint, remove direct identifiers from reports before running it, verify the appkey and base URL, and treat the output as health-management support rather than a medical diagnosis.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • System Prompt LeakageDirect Leakage, Indirect Extraction, Tool-Based Exfiltration
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (3)

Lp3

Medium
Category
MCP Least Privilege
Confidence
91% confidence
Finding
The skill declares no permissions, yet the documentation explicitly describes reading input files, optionally writing output files, and making network calls to an external/internal model endpoint using an appkey. This creates a capability/permission mismatch that can mislead reviewers and operators about the skill’s actual access needs, especially given it handles sensitive health data and transmits report contents off-host.

Missing User Warnings

Medium
Confidence
96% confidence
Finding
The skill sends full medical report text to a remote LLM service, which is highly sensitive personal health information, but it does not present any explicit privacy notice, consent checkpoint, or data-minimization step at the point of transmission. In a medical context, this materially increases the risk of unintended disclosure, regulatory non-compliance, and user harm if the remote service logs, retains, or mishandles the data.

Ssd 1

Medium
Confidence
95% confidence
Finding
The code inserts untrusted report text directly into the same user prompt as operational instructions, so a crafted report can include adversarial text that attempts to override formatting rules, suppress warnings, or manipulate the model's interpretation. In this medical-report setting, prompt injection can degrade output integrity and cause unsafe or misleading health guidance, even if it does not directly execute code.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.