Security audit

CreateVideo - Podcast to Video

Security checks for vulnerabilities and agentic risk

Overview

This is a disclosed podcast-video generation skill that uses ListenHub and ffmpeg, with privacy and setup considerations but no evidence of hidden or malicious behavior.

Install if you want this ListenHub-based podcast-video workflow. Use a revocable ListenHub API key, review the MCP package before installing it, avoid sending confidential scripts to the service, and confirm language, speaker IDs, template files, and output location before generating media.

SkillSpector

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (4)

Vague Triggers

Medium

Confidence: 92% confidence
Finding: The trigger description is broad enough that the skill may activate on loosely related requests about making videos from user copy, causing the agent to invoke media generation and third-party processing when the user did not clearly request this specific workflow. In this context, unintended activation increases the chance of surprise data handling, unnecessary file creation, and accidental transmission of user content to external services.

Missing User Warnings

Medium

Confidence: 97% confidence
Finding: The skill explicitly sends user-provided content to ListenHub MCP for podcast generation but does not warn users that their text will leave the local environment and be processed by a third-party service. Because this skill is designed to transform arbitrary user copy, it can easily handle sensitive personal, business, or confidential material, making undisclosed external transmission a meaningful privacy and compliance risk.

Natural-Language Policy Violations

Medium

Confidence: 83% confidence
Finding: Hard-coding the podcast language to Chinese without user choice can cause the system to transform content into an unintended language, which may misrepresent user intent, degrade output quality, or expose content to incorrect downstream processing. In a skill that may be triggered broadly and forwards content to an external service, lack of language consent also increases the chance of unexpected handling of multilingual or sensitive text.

Natural-Language Policy Violations

Medium

Confidence: 80% confidence
Finding: Setting Chinese as the default podcast language in configuration, without exposing a locale choice, creates a persistent mismatch risk for users whose content or audience is in another language. While not directly enabling code execution, it can still lead to unintended external processing behavior and content-generation outcomes that the user did not authorize.

Static analysis

No suspicious patterns detected.