ClawHub
Back to skill

Security audit

tech-evolution-analysis

Security checks across malware telemetry and agentic risk

Overview

The skill is mostly a coherent technology-analysis workflow, but it generates persistent HTML from loosely controlled content and one renderer allows inline HTML without sanitization.

Install only if you are comfortable with the skill writing local analysis folders and HTML reports. Treat generated HTML as active content: use trusted inputs, avoid copying raw untrusted web or database text into HTML fields, and sanitize or escape JSON/report content before sharing or opening reports in a browser context.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • 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
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (3)

Lp3

Medium
Category
MCP Least Privilege
Confidence
85% confidence
Finding
The skill instructs the agent to save multiple intermediate artifacts and generate HTML outputs, which implies file-write behavior despite no declared permissions. Undeclared write capabilities weaken reviewability and permission transparency, and could lead to unauthorized artifact creation or overwriting files in the agent workspace.

Tp4

High
Category
MCP Tool Poisoning
Confidence
73% confidence
Finding
The documented behavior goes beyond pure analysis and includes generating HTML visualizations and report artifacts, but these operational side effects are not clearly reflected in the declared purpose. Description-behavior mismatches are dangerous because reviewers and users may authorize a seemingly analytical skill without realizing it also transforms data into executable/renderable HTML outputs and persists derived files.

Intent-Code Divergence

Medium
Confidence
97% confidence
Finding
The renderer directly injects multiple JSON-controlled fields into HTML content without escaping, including title, sub, system names/tags, component names/tags, part names/tags, and footnotes. Because the docstring explicitly permits inline HTML, a malicious or untrusted JSON input can embed script-capable markup or event handlers, leading to stored XSS in the generated HTML when someone opens the file locally or serves it from a web context.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.