ClawHub
Back to skill

Security audit

unisound-chronic-screening

Security checks across malware telemetry and agentic risk

Overview

The skill does the stated health-screening task, but it sends sensitive health records to a remote LLM without implementing its promised de-identification and can save results despite claiming no local persistence.

Review this skill carefully before installing or using it with real resident records. Use only de-identified test data unless you trust the configured LLM endpoint and have an appropriate data-processing arrangement; avoid direct identifiers in input files, and treat --output files as sensitive medical records that need local access controls and retention handling.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • 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
70% confidence
Finding
Without declared permissions the skill's intent is opaque and cannot be validated.

Natural-Language Policy Violations

Low
Confidence
92% confidence
Finding
The natural-language content of the skill, including description, usage, and warnings, is presented only in Chinese. Under the policy, forcing a specific language without user opt-in can be a locale/language policy violation unless the restriction is explicitly justified or alternatives are offered.

Missing User Warnings

High
Confidence
94% confidence
Finding
The code sends resident health data to an HTTP API via the LLM call, which is a privacy-sensitive network transmission. Although the module docstring notes use of an internal medical model, there is no runtime warning, confirmation, or explicit disclosure near execution that the supplied health record will be sent to a remote service.

Missing User Warnings

Medium
Confidence
88% confidence
Finding
When --output is provided, the code writes chronic disease screening results derived from resident health data to a file. The write is operationally expected, but there is no warning or disclosure that potentially sensitive medical output will be persisted to local storage.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.