ClawHub
Back to skill

Security audit

一键加速视频下载

Security checks across malware telemetry and agentic risk

Overview

This skill is a disclosed video download workflow that asks for confirmation before downloading and clearly labels its optional translation-service data sharing.

Before installing, confirm you are comfortable with yt-dlp downloads from user-supplied public URLs, Chinese-localized default filenames, and hidden audio/subtitle preparation under the project folder. Use combined translation mode only if you agree to configure the listed API credentials locally and allow audio or subtitle text to be sent to the named external services.

SkillSpector

Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (3)

Description-Behavior Mismatch

Medium
Confidence
93% confidence
Finding
The combined-mode prompt expands a download skill into a broader workflow that requests credential setup and seeks consent for uploading audio and sending content to third-party ASR/translation services. That scope creep is security-relevant because a user invoking a download-only skill may not expect data exfiltration or credential configuration, increasing the chance of unintended disclosure or misuse. In this context, the prompt does at least disclose the external processing, which reduces severity, but the mismatch with the declared skill purpose still makes it a real vulnerability.

Vague Triggers

Medium
Confidence
89% confidence
Finding
The default prompt embeds broad activation logic and workflow control rules that can cause the agent to invoke the skill outside narrowly defined user intent. In a download-capable skill, ambiguous triggers increase the chance of unintended remote fetching, persistent session behavior, and hidden preparation steps being performed without clear confirmation boundaries.

Natural-Language Policy Violations

Medium
Confidence
93% confidence
Finding
The prompt forces translation of the original foreign title into Chinese and requires shared Chinese naming for downloaded assets and subtitle workflows without any user opt-in. This can override user expectations, alter source metadata deterministically, and create misleading or lossy filenames that affect provenance, auditability, and downstream processing.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.