ClawHub
Back to skill

Security audit

Video Add Content Cards

Security checks across malware telemetry and agentic risk

Overview

This skill is a disclosed video-editing helper that creates timed graphic overlays, though users should expect local file changes, tool execution, and some preview-time browser/network activity.

Install only if you are comfortable with the agent reading and writing the project’s video workflow files, running Python/ffmpeg/npx/Chrome-based rendering commands, and opening local browser review pages. For offline or privacy-sensitive work, review or bundle the example CDN and font dependencies before using the preview gallery.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (3)

Lp3

Medium
Category
MCP Least Privilege
Confidence
93% confidence
Finding
The skill instructs the agent to read and write project files and to execute local shell commands such as opening HTML files, invoking Python scripts, ffmpeg, and npx. Those capabilities are operationally relevant to the skill, but they are undeclared, which creates a permission-transparency gap: a user or orchestrator cannot accurately assess or constrain what the skill may do before execution. In this context the commands are mostly local and workflow-oriented rather than overtly exfiltrative, but hidden shell/file capabilities still increase risk if the skill or its dependencies are modified or abused.

Missing User Warnings

Low
Confidence
96% confidence
Finding
The file loads GSAP from cdnjs and fonts from Google Fonts at runtime, which causes outbound network requests to third parties and introduces supply-chain and privacy exposure. In this skill context, the HTML is intended for rendering video overlays, so hidden external fetches are unnecessary for core logic and can leak usage metadata or fail unpredictably in restricted/offline render environments.

Missing User Warnings

Medium
Confidence
95% confidence
Finding
The skill explicitly instructs the agent to execute native OS commands to open local files on the user's behalf, without requiring user confirmation at the point of execution. Even though the stated goal is benign UX convenience, agent-triggered command execution expands the attack surface and can normalize unsafe behavior, especially if file paths or targets ever become user-influenced or are swapped for unexpected resources.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.