Audio Command Executor

• Outcome: Review
• Latest audit: May 29, 2026 at 9:35 PM UTC
• Version:
  1.0.1

Overview

This skill has a clear audio-transcription purpose, but it needs review because it runs local tools on inbound file paths and treats audio-derived instructions like normal chat without clear safeguards.

Install only if the inbound audio directory is trusted and the runtime validates files stay inside that directory, uses safe argv-style command execution, restricts expected audio types, and asks before acting on sensitive commands spoken in audio.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• 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
• Excessive Agency: Unrestricted Tool Access, Autonomous Decision Making, Scope Creep

Findings (1)

Ssd 1

Medium

• Confidence: 97% confidence
• Finding: The skill instructs the agent to treat transcribed audio 'as normal chat input' and answer resulting questions or instructions without any trust boundary or safety filtering. Because audio is an untrusted external input, an attacker can embed spoken prompt-injection content that manipulates agent behavior, elicits sensitive information, or triggers unsafe downstream actions if the broader system grants tool access.

VirusTotal

65/65 vendors flagged this skill as clean.