Ai Video Editor Offline

Security checks across malware telemetry and agentic risk

Overview

This skill is marketed as offline local video editing, but its instructions use remote NemoVideo APIs, uploads, cloud GPU rendering, and download URLs.

Treat this as a cloud-based NemoVideo integration, not an offline editor. Only install or use it if you are comfortable sending videos, prompts, URLs, authentication-backed requests, and render jobs to NemoVideo, and avoid sensitive footage unless the publisher clearly documents consent, retention, deletion, and cancellation behavior.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (4)

Description-Behavior Mismatch

High
Confidence
98% confidence
Finding
The skill advertises offline/local processing and 'without uploading to the cloud,' but its setup and operation require obtaining a remote token, creating a remote session, uploading user media, and using cloud GPU rendering. This is a deceptive data-handling claim that can cause users to disclose sensitive local video content under false privacy expectations.

Intent-Code Divergence

High
Confidence
99% confidence
Finding
The documentation directly tells users exports occur 'without uploading to the cloud' while later sections describe cloud GPUs, remote upload endpoints, and download URLs. This contradiction materially misrepresents where user media is processed and stored, increasing the risk of privacy violations and non-compliant handling of sensitive recordings.

Description-Behavior Mismatch

High
Confidence
99% confidence
Finding
The implementation details clearly show remote uploads, SSE messaging, session-state polling, and rendered file download URLs, which directly contradict the marketed local/offline behavior. In context, this is especially dangerous because users may share confidential recordings, believing the data never leaves their device.

Missing User Warnings

High
Confidence
97% confidence
Finding
The skill both claims offline/local editing and instructs the agent to automatically connect to a processing API 'before doing anything else' without a clear warning. Automatically establishing remote sessions under misleading privacy claims undermines informed consent and may transmit metadata or prepare for uploads before the user understands cloud processing is involved.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal