ClawHub
Back to skill

Security audit

unisound-chief-complaint-diagnosis-inconsistent

Security checks across malware telemetry and agentic risk

Overview

This skill is a disclosed medical-record quality-check tool that sends limited record fields to a configured medical LLM and writes results locally, with privacy precautions users must follow.

Install only if you are allowed to process these medical records with the configured HiVoice MaaS or other selected endpoint. De-identify records before use, avoid --save-prepared unless debugging in a protected directory, set a finite timeout for production runs, and protect the app key and output files as sensitive healthcare data.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (4)

Lp3

Medium
Category
MCP Least Privilege
Confidence
95% confidence
Finding
The skill documentation describes capabilities to read local files, write output files, and send medical record text to an external MaaS endpoint, but it does not declare permissions or present explicit capability boundaries. In a medical-record processing context, undeclared file and network access is dangerous because it can lead to unreviewed exfiltration of sensitive health data and unexpected persistence of processed content.

Missing User Warnings

High
Confidence
94% confidence
Finding
The code sends chief complaint and diagnosis text from outpatient medical records to an external remote LLM endpoint. Even though the endpoint appears to be an internal/vendor medical model, this is still disclosure of sensitive health information to a third party service without visible consent handling, minimization, de-identification, or any warning in the implementation. In a medical context, this is especially dangerous because the transmitted text may contain protected health information and create compliance, privacy, and retention risks.

Missing User Warnings

Medium
Confidence
94% confidence
Finding
When --save-prepared is enabled, the script writes preprocessed medical record text directly to disk, which may contain highly sensitive PHI/PII. In a medical-record QC skill, this creates a real confidentiality risk because the data may be stored in an unprotected location, retained longer than intended, or accessed by unauthorized users or processes.

Missing User Warnings

Medium
Confidence
91% confidence
Finding
The script sends preprocessed medical record content to an LLM service via run_qc without any visible consent, disclosure, minimization, or endpoint restriction in this file. Because the input is outpatient medical record text, transmitting it to a model endpoint can expose PHI to network services, misconfigured hosts, or non-compliant backends, especially since the base URL is user-configurable.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.