Smart Router Intents

• Outcome: Pass
• Latest audit: May 30, 2026, 3:51 PM
• Version:
  1.0.1

Overview

This skill is a small local query classifier that recommends a model tier and does not itself contact external services or access private data.

Install this if you want local, low-stakes model-routing suggestions. For private, regulated, expensive, or safety-sensitive prompts, review the chosen route before sending the prompt to another model and avoid relying on this as an access-control or safety decision system.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access
• Supply Chain: Unpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (2)

Vague Triggers

Medium

• Confidence: 96% confidence
• Finding: The skill advertises itself as handling 'any query by intent,' which is overly broad and creates a high chance of unintended activation across ordinary user requests. In a routing skill, this is especially risky because misactivation can silently intercept prompts and redirect them to different models or capabilities, affecting cost, privacy, and behavior without clear user intent.

Vague Triggers

Medium

• Confidence: 98% confidence
• Finding: Defining the input trigger as 'any query' and then classifying into broad buckets without negative examples or hard boundaries makes the router applicable to nearly all conversations. This can cause overreach and prompt interception, where sensitive, safety-critical, or specialized requests are routed by a simplistic keyword scheme to inappropriate models or tools.

VirusTotal

66/66 vendors flagged this skill as clean.