Ai Subtitle Translator

• Outcome: Review
• Latest audit: May 29, 2026 at 9:33 PM UTC
• Version:
  1.0.0

Overview

This subtitle-translation skill uses a disclosed cloud video backend, but its instructions are broader than the advertised purpose and can route general editing requests to that backend.

Review before installing if you only want subtitle translation. This skill sends videos and prompts to NemoVideo cloud services, can create or use a NEMO_TOKEN, and may handle broader video-editing commands beyond subtitles. Avoid uploading private or confidential media unless you trust the service and understand its account, credit, retention, and deletion practices.

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 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 (5)

Description-Behavior Mismatch

Medium

• Confidence: 92% confidence
• Finding: The skill is presented as a subtitle translation tool, but the documented behavior expands into broader video editing, rendering, and media-processing actions. This scope drift can mislead users and host agents about what capabilities are being invoked, increasing the chance of unexpected data transfer or unintended execution of higher-impact operations.

Intent-Code Divergence

Medium

• Confidence: 95% confidence
• Finding: Although branded as subtitle translation, the skill explicitly routes unrelated requests like BGM, overlays, and aspect-ratio changes into backend actions. That mismatch broadens the operational scope beyond user expectations and can cause the agent to perform edits the user did not knowingly authorize.

Vague Triggers

Medium

• Confidence: 90% confidence
• Finding: The starter phrases are very broad, such as 'translate my video files,' and invite activation from ordinary conversational language. Overly permissive invocation patterns can cause accidental skill triggering, leading to unintended upload, processing, or session creation against a third-party backend.

Vague Triggers

Medium

• Confidence: 97% confidence
• Finding: The routing table contains an unbounded catch-all rule that sends 'Everything else' to the SSE backend. This is dangerous because arbitrary user text may be forwarded to a remote service and interpreted as editing commands, causing unintended actions, privacy exposure, and hard-to-predict behavior.

Missing User Warnings

Medium

• Confidence: 96% confidence
• Finding: The skill documentation does not clearly warn users up front that uploaded media, prompts, and session data are sent to a cloud backend. In a media-processing skill, that omission is significant because users may share sensitive videos or audio without understanding the data leaves the local environment.

VirusTotal

66/66 vendors flagged this skill as clean.