Remotion + Excalidraw + TTS

• Outcome: Review
• Latest audit: May 28, 2026, 9:54 AM
• Version:
  1.1.0

Overview

The skill is mostly purpose-aligned for making narrated videos, but it needs review because cloud TTS mode can expose API keys and voiceover text in command logs.

Review before installing if your voiceover text, diagrams, or API keys are sensitive. Use the default say backend or --voiceover-mp3 for local-only operation, avoid cloud TTS for confidential scripts, and be aware that npm dependencies are installed during rendering. If using OpenAI or ElevenLabs, do not run this in logs or CI output that other people can read unless the script is changed to redact credentials and payloads.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Supply Chain: Unpinned Dependencies, External Script Fetching, Obfuscated Code
• MCP Tool Poisoning: Hidden Instructions, Unicode Deception, Parameter Description Injection
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access

Findings (4)

Description-Behavior Mismatch

Medium

• Confidence: 92% confidence
• Finding: The manifest presents the skill as using local macOS `say`, but the documentation also advertises OpenAI and ElevenLabs modes that require external API access. This discrepancy can cause sensitive voiceover text or related project data to be sent off-host unexpectedly, especially in environments where 'local TTS' is relied upon for privacy or compliance.

Intent-Code Divergence

Medium

• Confidence: 90% confidence
• Finding: Labeling the skill as 'Local' while documenting remote TTS providers creates a misleading trust signal. In security-sensitive contexts, users may deliberately select 'local' tooling to avoid data exfiltration, so this wording increases the risk of accidental use of cloud services and underestimation of network exposure.

Missing User Warnings

Medium

• Confidence: 95% confidence
• Finding: The documentation mentions that the script will run `npm i` in a temporary project, but it does not elevate this as a clear security warning despite the fact that dependency installation executes network activity and introduces software supply-chain risk. This is dangerous because package installation can download untrusted code, trigger lifecycle scripts, and produce materially different security outcomes than a purely local render pipeline.

Missing User Warnings

Medium

• Confidence: 92% confidence
• Finding: When the openai or elevenlabs backend is selected, the script sends the full voiceover text to third-party APIs, but the CLI interface does not prominently warn the user that local content will leave the machine. In a video-generation skill, voiceover text may contain confidential prompts, internal architecture details, or regulated data, so silent transmission increases data-leak risk.

VirusTotal

66/66 vendors flagged this skill as clean.