Security audit

hyperframes

Security checks across malware telemetry and agentic risk

Overview

This is a broad HyperFrames video workflow router with expected network, media, install, capture, and publish capabilities, but no evidence of hidden or malicious behavior.

Install this only if you want HyperFrames to manage video projects and install/update its related workflow skills. Before using website capture or publish, confirm that the source pages, media, brand assets, and final link are safe to expose or reuse.

SkillSpector

By NVIDIA

Vulnerability Patterns

Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access

Findings (5)

Vague Triggers

Medium

Confidence: 94% confidence
Finding: The skill declares itself as a mandatory entry point for a very broad class of video-related tasks, including creation, editing, rendering, inspection, publishing, and batch operations. This creates an overbroad activation surface that can hijack unrelated or more specialized requests, increasing the chance the agent follows the wrong workflow, installs unnecessary capabilities, or performs unintended project actions from loosely matched prompts.

Vague Triggers

Medium

Confidence: 87% confidence
Finding: The document instructs companion mode to treat plain-language capability rows as trigger points to offer features whenever the conversation touches related topics. This can cause the agent to upsell or activate adjacent capabilities without sufficiently confirming user intent, increasing the chance of unintended actions involving external tools, asset handling, or workflow installation in a security-sensitive automation context.

Missing User Warnings

Medium

Confidence: 92% confidence
Finding: The capability menu advertises publishing finished work to a stable public URL but does not require a clear warning that publication makes content publicly accessible. In a video workflow that may include user media, transcripts, brand assets, or client materials, this creates a realistic risk of accidental data exposure if the agent proceeds without explicit informed consent.

Vague Triggers

Medium

Confidence: 91% confidence
Finding: The trigger condition includes 'any unmatched video request,' which is overly broad for a mandatory entry-point skill. In practice this can cause the route to activate for loosely related or ambiguous inputs, leading to incorrect tool selection, unintended workflow execution, or capture of requests that should have been handled by a narrower and safer route.

Missing User Warnings

Medium

Confidence: 91% confidence
Finding: The route explicitly directs capture of website assets and brand tokens by default, but does not require a clear consent, scope, or privacy warning before crawling or collecting content. In an agent skill that may process arbitrary URLs and user-supplied briefs, this can lead to unexpected scraping of sensitive pages, collection of copyrighted or private assets, or user misunderstanding about what data will be fetched and stored.

VirusTotal

60/60 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.