视频日语字幕

• Outcome: Review
• Latest audit: May 28, 2026 · 7:03 AM UTC
• Version: 1.0.0

Overview

This subtitle tool appears to do what it says, but it ships an embedded gateway token and can send video transcript text to translation services without clear user control.

Review before installing. Replace and rotate the embedded QClaw token, configure your own scoped credential outside the source, edit VIDEO_DIR and OUTPUT_DIR to dedicated folders, and only run it on videos whose audio/transcripts you are comfortable sending to QClaw and possibly MyMemory.

SkillSpector

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• 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
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access

Findings (6)

Lp3

Medium

• Category: MCP Least Privilege
• Confidence: 91% confidence
• Finding: The skill advertises and documents capabilities that require filesystem access, environment manipulation, shell execution, and network access, but it does not declare permissions. This undermines informed consent and sandboxing decisions, because a reviewer or runtime may treat it as lower risk than it really is while it can process local files and contact external services.

Tp4

High

• Category: MCP Tool Poisoning
• Confidence: 97% confidence
• Finding: The documented behavior goes beyond simple subtitle generation: it sends transcript content to external translation services, mentions a gateway token, and appears to batch-process a fixed local directory rather than only user-specified input. That creates significant privacy and security risk because sensitive video/audio-derived text may be exfiltrated and local files may be processed unexpectedly without clear user authorization.

Missing User Warnings

Medium

• Confidence: 95% confidence
• Finding: The description does not warn users that transcript or subtitle content may be sent to external services for translation. In this context, subtitle text is derived from video audio and can contain private conversations, copyrighted material, or confidential business content, so undisclosed transmission materially increases privacy and compliance risk.

Missing User Warnings

Medium

• Confidence: 94% confidence
• Finding: The workflow explicitly sends subtitle text to a local LLM gateway and, on failure, falls back to MyMemory API, but it does not warn that spoken content extracted from user videos may be transmitted to another process or an external third-party service. Subtitle text can contain sensitive personal, business, or copyrighted material, so silent off-device transmission creates a real privacy and data-handling risk.

Missing User Warnings

High

• Confidence: 99% confidence
• Finding: The script embeds a bearer token directly in source code, which creates an immediate secret exposure risk through source control, logs, backups, and reuse by unauthorized parties. In a skill context that may be shared, exported, or inspected by others, hardcoded credentials are especially dangerous because they enable silent abuse of the translation gateway and possible access to associated services.

Missing User Warnings

High

• Confidence: 98% confidence
• Finding: The code sends subtitle text derived from user videos to a local LLM gateway and, on failure, to an external MyMemory service without explicit notice or consent. Subtitle text can contain sensitive personal, confidential, or copyrighted material, so undisclosed transmission to third-party services creates a real privacy and data-governance risk that is amplified by this skill's purpose of processing user media.

Static analysis

No suspicious patterns detected.