Security audit

Code Analyze Tool Free

Security checks across malware telemetry and agentic risk

Overview

This skill appears to be a structured analysis prompt, but it asks for command execution authority without clearly limiting when or how commands may run.

Review this skill before installing. It does not show malicious behavior, but it grants shell command capability for a task that mostly appears to need reading and structured reasoning. Use it only in an environment where command execution is acceptable, and prefer a version that removes exec or documents an explicit allowlist of permitted commands.

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)

Description-Behavior Mismatch

Medium

Confidence: 94% confidence
Finding: The skill claims it is local/offline and that no network is required, yet it also permits command execution and explicitly suggests using network diagnostics like ping. This mismatch can mislead users and host agents about the real operational and trust boundary, increasing the chance that shell commands with network effects are invoked under a false assumption of safety.

Context-Inappropriate Capability

Medium

Confidence: 89% confidence
Finding: The manifest grants exec to a skill presented primarily as a Markdown-based structured analysis tool, without a narrowly scoped justification. Unnecessary command-execution capability expands the attack surface because any prompt or downstream instruction interpreted by the agent could cause local command invocation beyond simple analysis.

Intent-Code Divergence

Medium

Confidence: 92% confidence
Finding: The documentation describes the tool as pure Markdown, local, and offline, but nearby sections introduce exec-based behavior and elsewhere recommend network testing. Such contradictory guidance obscures the skill's real privileges and execution model, which is dangerous because operators may enable or trust the skill more broadly than they otherwise would.

Vague Triggers

Medium

Confidence: 85% confidence
Finding: The description advertises analysis of arbitrary input with very broad applicability, but provides little boundary-setting or exclusion logic. In an agent ecosystem, overbroad trigger scope can cause the skill to activate in unintended contexts, including on sensitive content or tasks where its exec-enabled environment creates unnecessary risk.

Vague Triggers

Medium

Confidence: 87% confidence
Finding: The capability coverage section lists many generalized keywords rather than precise activation rules. This can lead to accidental invocation across a wide range of contexts, which is especially concerning because the skill has read and exec permissions and presents itself as safe for broad local analysis.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.