Security audit

监控器

Security checks across malware telemetry and agentic risk

Overview

This monitoring skill is not clearly malicious, but it requests broad command and file authority without enforceable limits.

Review this skill before installing. Use it only for clearly defined monitoring tasks, and require explicit confirmation before any command execution, file modification, API call, credential use, or scheduled/background activity. Do not give it broad access to production systems or secrets unless you have added your own controls around allowed commands and data exposure.

SkillSpector

By NVIDIA

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
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

Findings (5)

Context-Inappropriate Capability

Medium

Confidence: 89% confidence
Finding: The skill declares the exec tool even though the document does not narrowly justify why arbitrary shell execution is required for a monitoring skill. In an agent context, exposing command execution for broad user-defined monitoring creates a path to run destructive or system-changing commands under the guise of operational checks.

Intent-Code Divergence

Medium

Confidence: 95% confidence
Finding: The file claims command execution is limited to a whitelist and sandbox, but the skill definition provides no actual enforcement mechanism or verifiable restriction. This mismatch can mislead users and downstream agents into trusting dangerous behavior, enabling arbitrary command execution or privilege misuse.

Vague Triggers

Medium

Confidence: 84% confidence
Finding: The activation language is extremely broad, inviting the skill to be used for 'anything' and across wide operational scenarios without clear bounds. In combination with read and exec capabilities, this increases the chance that the agent will apply the skill in unsafe contexts and perform risky actions not intended by the user.

Vague Triggers

Medium

Confidence: 82% confidence
Finding: Repeating the claim that the skill can monitor 'anything' reinforces an unbounded operational scope. For a skill with execution capability, vague scope materially raises misuse risk because the agent may generalize this into permission to inspect or act on arbitrary systems and data sources.

Missing User Warnings

Medium

Confidence: 87% confidence
Finding: The skill advertises file handling, API integration, and command execution but does not present clear user-facing warnings about system modification, sensitive data exposure, network effects, or command risk. Users may invoke the skill without understanding that it can read files, contact external services, or execute impactful operations.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.