Whisper Transcriber

Security checks across malware telemetry and agentic risk

Overview

This is a coherent local speech-to-text skill; its main risks are disclosed model downloads, local package installation, and handling sensitive audio transcripts.

Install if you are comfortable with the setup script changing local packages and downloading models from Hugging Face. For sensitive or regulated audio, get consent before transcription, limit who can read transcripts, avoid unnecessary retention, and pre-download or verify models in restricted environments.

SkillSpector

By NVIDIA

Vulnerability Patterns

Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep

Findings (2)

Missing User Warnings

Medium

Confidence: 86% confidence
Finding: The example shows automatic processing of inbound voice messages without any privacy notice, consent flow, retention guidance, or warning about sensitive content. In a speech-to-text skill, inbound audio may contain personal data, secrets, or regulated information, so normalizing unattended transcription increases the risk of unauthorized collection and handling of user data.

Missing User Warnings

Medium

Confidence: 88% confidence
Finding: The script automatically transcribes inbound voice messages as soon as they are handled, with no consent prompt, visibility notice, or policy check for potentially sensitive audio. Even though processing appears local via whisper.cpp, this still creates a privacy risk because users may not expect automatic conversion of voice content into searchable text that can be displayed, stored, or further processed.

VirusTotal

66/66 vendors flagged this skill as clean.

View on VirusTotal