ClawHub
Back to skill

Security audit

unisound-initial-record

Security checks across malware telemetry and agentic risk

Overview

This medical-record skill has a coherent purpose, but it handles sensitive health dialogue and has a real disclosure conflict around local persistence.

Review before installing or using with real patient data. Only provide de-identified dialogue, confirm the remote LLM endpoint is approved for your data, avoid --save-prepared unless you intentionally want sensitive preprocessed content written to disk, and delete any generated debug/output files under your retention policy.

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
87% confidence
Finding
The skill documentation describes capabilities to read multiple local file types, optionally write output or prepared data to disk, and send medical dialogue to an external API, but no explicit permission declaration is shown. In a medical-record workflow, this mismatch is risky because users may not realize the skill can access local files, persist sensitive content, and transmit protected health information over the network.

Missing User Warnings

High
Confidence
97% confidence
Finding
The README exposes a `--save-prepared` option for debugging without clearly warning that it may write preprocessed medical dialogue to disk. Because the input contains highly sensitive health information, enabling local persistence can create privacy breaches, accidental retention, backup propagation, or unauthorized access long after the run completes.

Missing User Warnings

High
Confidence
98% confidence
Finding
The skill sends raw medical dialogue content to a remote LLM API, which likely contains highly sensitive personal and health information. In a medical context, transmitting PHI without explicit user notice, consent, minimization, and documented handling guarantees creates serious privacy and compliance risk if the endpoint is misconfigured, logged, retained, or operated outside approved boundaries.

Missing User Warnings

Medium
Confidence
96% confidence
Finding
When --save-prepared is used, the script writes preprocessed dialogue content to disk, which may include sensitive medical conversations and identifiers. In this healthcare context, local debug artifacts can become an unintended disclosure vector through weak filesystem permissions, backups, shared environments, or accidental retention.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.