Vlmrun Cli Skill

• Outcome: Pass
• Latest audit: May 28, 2026 at 11:00 AM UTC
• Version:
  0.1.1

Overview

This skill is a straightforward VLM Run CLI integration, with expected remote-service, API-key, package-install, and artifact-cache risks that users should understand before using it.

Before installing, verify that you trust the vlmrun package and service, use a dedicated VLM Run API key, and avoid sending confidential images, videos, PDFs, contracts, invoices, meeting recordings, or prompts unless you are comfortable with external processing and local artifact caching.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access
• Supply Chain: Unpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (3)

Vague Triggers

Medium

• Confidence: 89% confidence
• Finding: The skill metadata advertises very broad trigger phrases such as 'generate an image/video', 'analyze this image/video', and 'process this PDF', which overlap with many ordinary user requests. This can cause the skill to activate unexpectedly and route user content to an external service without a clear, deliberate invocation boundary, increasing the risk of unintended data exposure and tool misuse.

Missing User Warnings

Medium

• Confidence: 94% confidence
• Finding: The skill is designed to upload images, videos, and documents to the VLM Run/Orion external service, but it does not prominently warn users that their files and prompts will be transmitted off-box. Because the covered inputs include sensitive media and PDFs, users may unknowingly send confidential or regulated data to a third party.

Missing User Warnings

Low

• Confidence: 91% confidence
• Finding: The documentation encourages session continuation and notes local artifact caching, but it does not warn that prior conversation context, uploaded inputs, and generated artifacts may persist locally and possibly remotely as part of the session. This can lead users to reuse sessions containing sensitive material without understanding the retention and cross-task exposure risks.

VirusTotal

63/63 vendors flagged this skill as clean.