Security audit

Evez Skill Vetter

Security checks across malware telemetry and agentic risk

Overview

This is a local skill-scanning helper with incomplete safety claims, but no evidence of hidden data sending or dangerous actions.

Use this as a lightweight first-pass scanner only. Run it on a copied or narrowly scoped skill directory, do not treat a low score as proof that a skill is safe, and expect to install or provide the click dependency yourself if it is missing.

SkillSpector

By NVIDIA

Vulnerability Patterns

MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
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 (3)

Description-Behavior Mismatch

Medium

Confidence: 97% confidence
Finding: The skill description promises permission auditing, but the code shown does not parse any manifest or permission declarations and therefore cannot verify requested capabilities. This can mislead operators into approving a skill under the assumption that permission scope was reviewed when it was not.

Description-Behavior Mismatch

Medium

Confidence: 97% confidence
Finding: The skill description promises permission auditing, but the code shown does not parse any manifest or permission declarations and therefore cannot verify requested capabilities. This can mislead operators into approving a skill under the assumption that permission scope was reviewed when it was not.

Description-Behavior Mismatch

Medium

Confidence: 94% confidence
Finding: The tool claims to detect data exfiltration risk, but its network detection is incomplete and mostly focused on a limited set of write calls and browser APIs. Many exfiltration paths such as requests.get, socket libraries, websocket clients, FTP/SMTP, DNS-based channels, and higher-level SDKs are not covered, so risky skills may be underreported.

VirusTotal

53/53 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

Detected: suspicious.dangerous_exec, suspicious.dynamic_code_execution

Shell command execution detected (child_process).

Critical

Code: suspicious.dangerous_exec
Location: scripts/vet.py:23

Dynamic code execution detected.

Critical

Code: suspicious.dynamic_code_execution
Location: scripts/vet.py:22