Security audit

dispatching-parallel

Security checks across malware telemetry and agentic risk

Overview

The skill is not malicious, but it asks for broad read/exec-capable automation while giving vague scope, credential guidance, and unsupported security assurances.

Review this skill carefully before installing. It has no visible malicious payload, but it should be tightened before use: limit activation to explicit parallel-agent dispatch tasks, require user approval before shell commands or broad file reads, document exactly how credentials are handled, and remove or substantiate the security claims.

SkillSpector

By NVIDIA

Vulnerability Patterns

Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access

Findings (5)

Intent-Code Divergence

Medium

Confidence: 92% confidence
Finding: The skill documentation makes concrete security assurances about encryption, access control, logging, monitoring, and auditing, but the file contains no implementation details or enforceable controls that substantiate those claims. In an agent ecosystem, this can mislead operators into sending sensitive data to a skill under false security assumptions, increasing the risk of inappropriate trust and unsafe deployment.

Intent-Code Divergence

Medium

Confidence: 89% confidence
Finding: The FAQ claims support for multithreading, asynchronous processing, API integration, and data encryption without corresponding implementation or operational detail in the skill file. These unsupported capability claims can cause users or orchestrators to rely on behaviors that do not exist, which is especially risky when the claims include security properties such as encryption.

Vague Triggers

High

Confidence: 86% confidence
Finding: The activation guidance is extremely broad, covering generic efficiency improvement, workflow automation, batch processing, and development/data-analysis scenarios, which could cause the skill to trigger for many unrelated requests. In agent routing contexts, overbroad matching can result in unintended invocation of a skill that has exec capability, expanding attack surface and increasing the chance of unsafe tool use in the wrong context.

Vague Triggers

Medium

Confidence: 78% confidence
Finding: The body text describes the skill's processing at a high level but does not clearly distinguish when it should versus should not activate beyond vague efficiency and automation themes. This ambiguity can cause accidental routing and misuse, especially because the skill advertises broad workflow capabilities and may be paired with powerful tools.

Natural-Language Policy Violations

Medium

Confidence: 74% confidence
Finding: The skill claims '支持中文交互' in a way that static analysis interpreted as enforced Chinese interaction without clear user choice. While this is not a classic exploit primitive, forcing or implicitly defaulting language behavior can degrade user comprehension, hide important warnings from non-Chinese users, and contribute to unsafe operation when prompts, outputs, or consent details are misunderstood.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.