Security audit

Flow Control Hub Free

Security checks across malware telemetry and agentic risk

Overview

This skill is a desktop automation guide whose core purpose is clear, but its examples can type into live windows, save screenshots, and capture credentials without enough safeguards.

Review this skill carefully before installing. Use it only in controlled desktop sessions, keep sensitive windows closed, avoid running it on login pages or financial/account-management screens, and inspect any generated script before it clicks, types, presses Enter, saves files, or captures screenshots. Treat saved screenshots and logs as sensitive files.

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 (6)

Description-Behavior Mismatch

Medium

Confidence: 93% confidence
Finding: The documentation claims all operations are local and do not involve cloud calls, yet it also describes platform-side LLM model routing. That mismatch can mislead users about where prompts, instructions, or automation context may be processed, creating a trust and data-handling risk when users rely on the skill for sensitive desktop workflows.

Intent-Code Divergence

Medium

Confidence: 89% confidence
Finding: The skill advertises boundary checking and automatic coordinate validation, but the examples do not implement such checks. In a desktop automation skill, missing guardrails can cause clicks, drags, or keystrokes to hit unintended UI targets, potentially triggering destructive actions in the active window.

Missing User Warnings

Medium

Confidence: 95% confidence
Finding: The screenshot features encourage capturing and saving screen contents without clearly warning that screenshots may include passwords, personal data, tokens, chats, or other sensitive information visible on screen. Because the images are saved locally, they can persist beyond the session and become an accidental disclosure channel.

Missing User Warnings

Medium

Confidence: 92% confidence
Finding: The skill promotes keyboard and mouse automation but does not clearly warn that actions apply to the currently focused window and may unintentionally submit forms, overwrite text, save files, send messages, or confirm destructive dialogs. In a live desktop environment, focus errors or timing drift can quickly turn routine automation into unintended side effects.

Ssd 3

Medium

Confidence: 96% confidence
Finding: The sample controller logs and stores full input text, which can include passwords, personal data, API keys, or business-sensitive content typed during automation. Persisting this data in logs and action history creates an unnecessary secondary exposure path independent of the target application.

Ssd 3

High

Confidence: 99% confidence
Finding: The workflow explicitly takes a screenshot after credentials are filled, which is a direct sensitive-data capture pattern. This can expose usernames, passwords, session hints, or adjacent account data in plain image files that may be copied, backed up, or reviewed by unauthorized parties.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.