Trimmer Js

Security checks across malware telemetry and agentic risk

Overview

This is an instruction-only cloud video editing connector, and its token use, uploads, rendering, and session state are disclosed enough to fit its purpose.

Install only if you are comfortable sending selected videos, audio/images if used, media URLs, and edit prompts to NemoVideo for cloud processing. Protect the NEMO_TOKEN, monitor credit use, and avoid confidential or sensitive media unless you trust the provider’s privacy and retention practices.

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)

Description-Behavior Mismatch

Medium

Confidence: 95% confidence
Finding: The manifest presents the skill as a simple video trimmer, but the documented behavior exposes a broader cloud media-editing and generation workflow, including stateful editing, rendering, and track manipulation. This mismatch can mislead users and host platforms about the true capability and data handling scope, weakening consent and policy enforcement.

Context-Inappropriate Capability

Medium

Confidence: 93% confidence
Finding: Accepting remote media by URL introduces a larger trust boundary than local file trimming and can be abused to fetch unexpected or sensitive resources if the backend follows arbitrary URLs. Even if the fetch occurs server-side, this expands the attack surface and is not justified by the narrowly described trimming purpose.

Intent-Code Divergence

Medium

Confidence: 89% confidence
Finding: The skill claims support for a narrow set of video formats and size limits, but later documents many more media types including audio and images. This inconsistency can cause reviewers and users to underestimate the skill's real input surface, which matters for security review, data handling, and abuse prevention.

Vague Triggers

Medium

Confidence: 96% confidence
Finding: Routing nearly all unmatched prompts to a generic SSE backend creates an overly broad execution path where arbitrary user input is forwarded to a remote agent-like service. In this skill's context, that can enable unintended operations beyond trimming, increase prompt-injection exposure, and bypass the principle of least privilege.

Missing User Warnings

Medium

Confidence: 94% confidence
Finding: The skill encourages users to upload media but does not clearly warn them up front that files and prompts are sent to a third-party cloud backend for processing. For user-provided video content, this is a meaningful privacy and consent issue because clips may contain sensitive personal or proprietary data.

VirusTotal

No VirusTotal findings

View on VirusTotal