Security audit

Learn Tool Free

Security checks across malware telemetry and agentic risk

Overview

This learning skill mostly provides study-plan guidance, but it asks for file-writing and command-running authority without clear limits or user controls.

Review before installing. It does not show exfiltration, destructive scripts, or hidden code, but its requested command and file-writing powers are not well scoped for a study assistant. Install only if you are comfortable supervising any commands, package installs, callbacks, file writes, or cleanup actions it may suggest.

SkillSpector

By NVIDIA

Vulnerability Patterns

Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
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)

Context-Inappropriate Capability

Medium

Confidence: 90% confidence
Finding: The skill declares `exec` and `write` despite its stated purpose being a learning assistant that should primarily generate plans, exercises, and assessments. Granting shell-capable functionality to a low-risk educational workflow unnecessarily expands the attack surface and could let a prompt trigger local command execution or filesystem changes unrelated to the user’s learning task.

Context-Inappropriate Capability

Medium

Confidence: 84% confidence
Finding: The documentation advertises generic create/query/export/modify/delete capabilities and runtime configuration patterns that are far broader than the described educational use case. This kind of vague, overpowered operational scope can be abused to justify unintended actions, including manipulating files or invoking tooling beyond generating study content.

Intent-Code Divergence

Low

Confidence: 72% confidence
Finding: The skill presents itself as pure Markdown requiring no API key, but elsewhere declares execution capability and includes installation of `requests`, creating a misleading trust signal about what the skill can actually do. This inconsistency can cause users or agents to underestimate the possibility of external communication or local code execution.

Vague Triggers

Medium

Confidence: 79% confidence
Finding: The trigger examples and usage scope are broad natural-language prompts without clear activation boundaries, permission checks, or limits on tool use. In an agent setting, ambiguous invocation rules increase the chance that unrelated user input could activate the skill and chain into write/exec-enabled behavior.

Missing User Warnings

Medium

Confidence: 93% confidence
Finding: The skill exposes `write` and `exec` capabilities in its Markdown metadata but gives users no warning that it may alter files or run system commands. For a learning assistant, that omission is dangerous because users would reasonably expect content generation, not system-impacting actions, making accidental or adversarial misuse more likely.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.