Security audit

Network Tool

Security checks across malware telemetry and agentic risk

Overview

This network skill is not destructive, but it exposes broader active network probing and arbitrary HTTP request abilities than its description clearly tells users.

Review before installing. Use this only where outbound network tests and port scanning are authorized, and avoid running its curl or port-scan commands against systems you do not own or administer. The main issue is under-disclosure and broad network authority, not evidence of credential theft, persistence, or destructive behavior.

SkillSpector

By NVIDIA

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access

Findings (6)

Lp3

Medium

Category: MCP Least Privilege
Confidence: 90% confidence
Finding: The skill advertises and appears to enable shell and network-capable operations, but declares no permissions. That creates a transparency and control gap: users or platforms may authorize and trust the skill without understanding that it can execute network and shell-level actions, increasing the chance of misuse or unintended system/network access.

Tp4

High

Category: MCP Tool Poisoning
Confidence: 84% confidence
Finding: The described purpose is local network diagnostics and configuration analysis, but the detected behaviors include arbitrary outbound HTTP requests, port scanning, remote DNS/public-IP queries, and speed-test style downloads. This mismatch is dangerous because it expands the skill from passive inspection into active external network interaction that could be abused for reconnaissance, data exfiltration, or unauthorized probing while appearing benign to the user.

Description-Behavior Mismatch

Medium

Confidence: 88% confidence
Finding: The curl_request function turns a diagnostic skill into a general-purpose HTTP client capable of sending arbitrary requests, headers, and bodies to attacker-chosen destinations. In an agent setting, this materially expands the skill's power and can enable SSRF-style access to internal services, data exfiltration, or unintended interactions with external systems.

Description-Behavior Mismatch

Medium

Confidence: 90% confidence
Finding: Active port-range scanning exceeds passive network diagnostics and provides reconnaissance capability against arbitrary hosts. In an agent environment, this can be abused to probe internal networks and map reachable services, increasing the impact of prompt injection or tool misuse.

Missing User Warnings

Medium

Confidence: 93% confidence
Finding: Public IP lookup contacts third-party services, disclosing the system's IP address and metadata such as timing and user agent to external providers without an explicit warning or consent flow. In privacy-sensitive or enterprise environments, that outbound traffic may violate expectations or policy.

Missing User Warnings

Medium

Confidence: 94% confidence
Finding: The speed test downloads external content from a third-party host without explicit disclosure about bandwidth consumption or external connectivity. This can create privacy, cost, and policy issues, especially in constrained or monitored environments.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.