Security audit

AI Data Visualizer

Security checks across malware telemetry and agentic risk

Overview

This skill mostly matches its data-visualization purpose, but generated dashboards can turn crafted CSV/JSON content into active browser code when opened.

Review before installing. Use it only with trusted datasets or in a restricted workspace, avoid opening dashboards generated from unknown CSV/JSON files, and treat the output HTML as sensitive because it can contain raw data. A safer version should escape/render all user data as text, document temporary-file cleanup, and offer a local or integrity-pinned Chart.js option.

SkillSpector

By NVIDIA

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
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

Findings (7)

Lp3

Medium

Category: MCP Least Privilege
Confidence: 88% confidence
Finding: The skill documentation indicates file-writing behavior by generating dashboards and instructing users to write direct input into a temporary CSV file, yet it declares no permissions. This creates a transparency and governance gap: users and hosting platforms may not realize the skill can persist data to disk, including potentially sensitive user-supplied datasets.

Description-Behavior Mismatch

Medium

Confidence: 94% confidence
Finding: The generated dashboard pulls Chart.js from a third-party CDN, which creates an external network dependency that contradicts the expectation of a local/offline data visualization tool. Opening the generated HTML leaks access metadata to the CDN and exposes users to supply-chain risk if the remote script is tampered with or blocked.

Vague Triggers

Medium

Confidence: 82% confidence
Finding: The trigger phrases are broad enough to activate on common user requests like 'plot' or 'chart,' which can cause the skill to engage unexpectedly outside a clearly consented workflow. In an agent setting, overbroad activation increases the chance of processing unrelated or sensitive data and invoking file-generation behavior when the user did not specifically request this skill.

Missing User Warnings

Medium

Confidence: 92% confidence
Finding: The workflow says direct user data content may be written to a temporary CSV file, but this is not surfaced as a user-facing warning in the skill description. That omission matters because uploaded or pasted datasets may contain sensitive information, and silently persisting them to disk increases exposure through leftover temp files, backups, or other local access.

Missing User Warnings

Medium

Confidence: 90% confidence
Finding: The generated HTML loads Chart.js from a CDN, but the skill does not warn users that opening the dashboard may require network access and may expose metadata such as IP address, user agent, referrer, or access timing to a third party. This is especially relevant when dashboards are built from sensitive internal data, because users may assume the output is entirely local and offline-safe.

Missing User Warnings

Medium

Confidence: 89% confidence
Finding: The script serializes and embeds raw input rows directly into the generated HTML dashboard, which can expose sensitive dataset contents to anyone who gains access to the file. In this skill's context, users are likely to visualize internal CSV/JSON business data, so automatically including raw records increases the risk of unintended disclosure.

Missing User Warnings

Medium

Confidence: 95% confidence
Finding: Loading Chart.js from a remote CDN means that opening the generated local HTML causes a network request without clear disclosure to the user. This can reveal that the dashboard was opened and can transmit metadata such as IP address, user agent, referrer/file context behavior, and timing to a third party.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.