Security audit

data-viz-wizard

Security checks across malware telemetry and agentic risk

Overview

This CSV chart skill is useful, but its generated reports run browser JavaScript, load Chart.js from a third-party CDN, and do not safely escape some CSV-derived text.

Review before installing. Use this only with trusted CSV files or after adding HTML escaping/sanitization, and avoid opening generated reports with sensitive data in environments where third-party network requests are not allowed. A safer version would bundle Chart.js locally, add SRI/CSP protections if a CDN is retained, and clearly warn that generated reports are active HTML.

SkillSpector

By NVIDIA

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access

Findings (6)

Lp3

Medium

Category: MCP Least Privilege
Confidence: 89% confidence
Finding: The skill advertises command-line workflows that read CSV input and write generated HTML output, which are code capabilities, yet no permissions are declared in the skill metadata. This creates a transparency and policy-enforcement gap: hosts or reviewers may assume the skill is passive documentation while it actually performs filesystem operations.

Context-Inappropriate Capability

Medium

Confidence: 95% confidence
Finding: The generated HTML unconditionally loads Chart.js from a public CDN, causing a local CSV-to-HTML workflow to make external network requests when the output is opened. This creates privacy and supply-chain risk: viewing the chart leaks access metadata to the CDN and makes rendering dependent on remote third-party code that could change or be unavailable.

Context-Inappropriate Capability

Medium

Confidence: 95% confidence
Finding: The dashboard template also imports Chart.js from an external CDN, adding network behavior to a tool whose expected function is local processing of CSV data into HTML. This broadens the trust boundary to a third party and can expose user environment details or fail unpredictably if the CDN is blocked, compromised, or modified.

Missing User Warnings

Medium

Confidence: 87% confidence
Finding: The README explicitly promotes generation of standalone HTML files and tells users to open them in a browser, but it does not warn that the output necessarily contains executable JavaScript. If untrusted CSV content is embedded into the generated page without robust escaping in the implementation, users may open a file that executes attacker-controlled script in the browser, making the documentation materially unsafe by omission.

Missing User Warnings

Medium

Confidence: 92% confidence
Finding: The single-chart HTML performs a third-party network fetch without any user-facing warning, so users may believe the output is safe to open offline when it is not. In a data-visualization skill handling potentially sensitive local CSVs, undisclosed external dependencies increase the risk of privacy leakage and violate user expectations around local-only processing.

Missing User Warnings

Medium

Confidence: 92% confidence
Finding: The dashboard output similarly fetches Chart.js from a public CDN without warning the user, concealing network activity in what appears to be a local reporting artifact. That hidden behavior is especially problematic because dashboards may be opened in restricted or sensitive environments where any outbound request is unexpected or disallowed.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.