Security audit

R ggplot Quickplot

Security checks across malware telemetry and agentic risk

Overview

This appears to be a CSV-to-chart helper with disclosed dependency setup, not hidden data theft or destructive behavior.

Install only if you are comfortable with a plotting helper that may install ggplot2 from CRAN when run locally. In controlled or offline environments, preinstall or pin R dependencies and avoid the Docker publishing path unless you intend to push an image. Also verify the missing run_plot.R, Dockerfile, and Singularity.def from a trusted source before relying on the skill.

SkillSpector

By NVIDIA

Vulnerability Patterns

Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
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

Findings (5)

Description-Behavior Mismatch

Medium

Confidence: 90% confidence
Finding: The launcher silently expands its behavior from local CSV plotting to installing software from the network. That introduces supply-chain and execution risk, because package installation runs remote package install logic and may fetch unpinned content without the user's explicit consent.

Context-Inappropriate Capability

Medium

Confidence: 86% confidence
Finding: For a CSV-to-chart helper, automatic package installation is beyond the minimum expected capability and increases attack surface. In this skill context, users likely expect visualization only, not network access and environment modification, so the behavior is more dangerous because it is surprising and implicit.

Description-Behavior Mismatch

Medium

Confidence: 94% confidence
Finding: The script performs implicit network-based package installation at runtime via install.packages('ggplot2'), which expands the skill from local CSV plotting into remote code/package retrieval. This introduces supply-chain and reproducibility risk, especially in automated or privileged environments where execution of newly downloaded package code may occur without explicit user consent.

Context-Inappropriate Capability

Medium

Confidence: 91% confidence
Finding: For a chart-generation skill, silently installing external packages is unnecessary and creates an unjustified capability to fetch and execute third-party content. In security-sensitive environments, this can bypass expectations that the skill only processes local files and can enable supply-chain compromise or unexpected outbound network access.

Missing User Warnings

Medium

Confidence: 93% confidence
Finding: The script installs an R package over the network without warning or confirmation, which can lead to unexpected downloads, system changes, and exposure to repository or dependency compromise. Even if CRAN is generally trusted, silent network installation is unsafe in a user-facing launcher because it removes informed consent and auditability.

VirusTotal

66/66 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.