ClawHub
Back to skill

Security audit

unisound-med-content-generation

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed medical LLM helper that sends user-provided medical text to a configured model endpoint and can save results, so users should handle patient data carefully.

Install only if you are comfortable sending the provided medical text to the configured model API. Do not submit identifiable patient information unless your organization has approved that endpoint and workflow; prefer de-identified inputs, protect the app key, and avoid --output on shared filesystems unless the saved JSON is allowed to contain the original question and metadata.

SkillSpector

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 (3)

Lp3

Medium
Category
MCP Least Privilege
Confidence
91% confidence
Finding
The skill documentation describes code behavior that reads local files/stdin, writes output files, and sends prompts and potentially document contents to a remote model API, but it declares no corresponding permissions. This creates a real security issue because integrators or users may invoke the skill without realizing it can exfiltrate sensitive medical content over the network and persist data to disk, especially given support for clinical summaries and record generation.

Missing User Warnings

High
Confidence
97% confidence
Finding
The skill sends user-supplied medical questions directly to a remote third-party API and the question content may contain PHI/PII, clinical details, or other sensitive medical data. There is no consent gate, redaction step, or prominent disclosure in the program flow, so operators can unintentionally exfiltrate regulated health information outside the local environment.

Missing User Warnings

Medium
Confidence
91% confidence
Finding
The tool writes prompts, metadata, and model answers to stdout and optional output files, which can persist sensitive medical content and identifiers in logs, shell history pipelines, shared filesystems, or downstream tooling. Because the output object includes question and meta fields verbatim, sensitive content may be stored without minimization or warning.

Static analysis

No suspicious patterns detected.