视频去除水印

• Outcome: Review
• Latest audit: May 29, 2026 at 9:34 PM UTC
• Version:
  1.0.2

Overview

The visible files do not show malware, but the skill asks users to run an unverified local binary and optional HTTP service that are not adequately scoped or documented.

Install only if you trust the publisher and can independently verify the parser binary source. Avoid private or token-bearing share links, and do not expose the optional HTTP service beyond localhost unless you understand and control its network access.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• MCP Least Privilege: Underdeclared Capability, Wildcard Permission, Missing Permission Declaration
• MCP Tool Poisoning: Hidden Instructions, Unicode Deception, Parameter Description Injection
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands

Findings (6)

Lp3

Medium

• Category: MCP Least Privilege
• Confidence: 88% confidence
• Finding: The skill invokes shell scripts and native binaries but does not declare corresponding permissions or execution capabilities. Undeclared code execution increases audit difficulty and can expose users to unexpected local command execution, especially when handling untrusted URLs as input.

Tp4

High

• Category: MCP Tool Poisoning
• Confidence: 84% confidence
• Finding: The skill is presented primarily as a one-shot video parsing tool, but it also exposes persistent HTTP server functionality with a Web UI. Expanding from local parsing to a service changes the attack surface significantly, creating opportunities for unintended exposure, local network access, or unsafe request handling not disclosed in the stated purpose.

Intent-Code Divergence

Medium

• Confidence: 86% confidence
• Finding: The README claims the tool is 'only local' and 'safe/reliable' because it uses a local executable, but later documents starting an HTTP service. That mismatch can mislead users into underestimating exposure, especially if the service binds beyond localhost or handles untrusted URLs and content.

Missing User Warnings

Medium

• Confidence: 91% confidence
• Finding: The README instructs users to start an HTTP service without any warning about who can reach it, what data it processes, or the risks of exposing a video-parsing endpoint. In this skill context, the service likely accepts attacker-controlled URLs, making undocumented network exposure more dangerous because it could enable abuse, unintended data access, or SSRF-like behavior depending on implementation.

Vague Triggers

Medium

• Confidence: 74% confidence
• Finding: The trigger phrases are broad enough to match generic requests like downloading or parsing videos beyond the narrowly intended use case. Overbroad activation can cause the skill to run in contexts the user did not intend, potentially sending third-party links to local tooling or services without clear consent.

Missing User Warnings

Medium

• Confidence: 80% confidence
• Finding: The skill asks users to submit third-party share links for processing but does not clearly warn that those links and associated metadata may be disclosed to local services, scripts, or remote platforms during resolution. This can surprise users and create privacy issues, especially when links contain identifiers, tracking parameters, or access-scoped tokens.

VirusTotal

64/64 vendors flagged this skill as clean.