video-transcript-pro

• Outcome: Pass
• Latest audit: May 29, 2026 at 8:28 PM UTC
• Version:
  2.3.0

Overview

This appears to be a legitimate transcription and content-drafting skill, but users should be careful with private recordings because it can automatically use web search and remember preferences.

Install only if you are comfortable with automated transcription, local output file creation, and possible online searches based on transcript content. For private, regulated, legal, medical, business, or unpublished recordings, tell the agent to avoid web_search/web_fetch, confirm the output directory before processing, and inspect or clear MEMORY.md if you do not want preferences reused.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Excessive Agency: Unrestricted Tool Access, Autonomous Decision Making, Scope Creep
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access
• Supply Chain: Unpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (9)

Context-Inappropriate Capability

Medium

• Confidence: 96% confidence
• Finding: The skill explicitly reads and writes persistent user preferences in `MEMORY.md` and uses them to drive future inferred behavior, but it does not describe consent, retention limits, access controls, or a narrowly scoped purpose. This creates a privacy and autonomy risk because personal preferences can be silently accumulated and reused across sessions without the user's explicit approval each time.

Context-Inappropriate Capability

Medium

• Confidence: 93% confidence
• Finding: The skill automatically performs `web_search`/`web_fetch` on transcript content to supplement terminology, which introduces external network egress beyond the core local transcription workflow. If transcript content contains confidential, proprietary, or personal information, this behavior can leak sensitive context to third-party services without user awareness.

Missing User Warnings

Medium

• Confidence: 94% confidence
• Finding: The README explicitly advertises automatic online knowledge searches but does not warn that transcript content, filenames, or surrounding context may be sent to external services. For a transcription workflow, this can expose sensitive audio-derived data, proprietary material, or personal information without meaningful user awareness or consent.

Missing User Warnings

Medium

• Confidence: 91% confidence
• Finding: The skill says it will auto-continue after 3 minutes and generate multiple output files, but does not clearly warn users that processing and file creation may happen without final confirmation. In practice, this can lead to unintended handling of sensitive media and unwanted persistence of derived content on disk.

Missing User Warnings

Medium

• Confidence: 95% confidence
• Finding: Stating that online searches occur automatically and without confirmation creates a clear risk of unapproved network access and external disclosure of transcript-derived context. This is especially sensitive in transcription workflows, where audio may contain confidential business, legal, medical, or personal information.

Missing User Warnings

Medium

• Confidence: 97% confidence
• Finding: The skill instructs the agent to store user preferences in `MEMORY.md` for future use without any visible privacy notice, consent flow, or user control. Persistent storage of behavioral preferences can expose personal data and normalize silent profiling if the file is later accessed, reused, or retained indefinitely.

Missing User Warnings

Medium

• Confidence: 95% confidence
• Finding: Automatic web searches on transcript content can disclose parts of the user's audio/video content to external providers, yet the skill does not warn the user or obtain consent for that disclosure. In a transcription workflow, uploaded material may contain confidential meeting content, personal information, or unpublished intellectual property, which makes silent external transmission risky.

Ssd 3

Medium

• Confidence: 97% confidence
• Finding: Persisting user preferences for future automatic reuse without explicit per-use consent creates a durable behavior-modification channel that can affect later interactions without the user's active awareness. This is dangerous because stored preferences may become stale, overbroad, or privacy-sensitive, and can steer future outputs in ways the user did not intend for the current task.

Ssd 3

Medium

• Confidence: 95% confidence
• Finding: Using prior history from `MEMORY.md` as an input to infer topic, audience, style, and automatic continuation means historical user data directly influences current task execution. In context, this makes the workflow more dangerous because the skill already auto-continues after timeout, so hidden memory can silently shape outputs and decisions without a fresh user confirmation.

VirusTotal

66/66 vendors flagged this skill as clean.