ClawHub

Model Router

Security checks across malware telemetry and agentic risk

Overview

This is an instruction-only model-routing skill whose ability to choose cheaper or premium AI models is disclosed and consistent with its purpose.

Safe to install as a routing guide, with normal caution: define a maximum spend or preferred model tier, ask for approval before premium or gateway routing on sensitive work, and avoid auto-routing private prompts through providers you do not trust.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (2)

Vague Triggers

Medium
Confidence
92% confidence
Finding
The skill description includes broad trigger phrases such as 'which model should I use', 'optimize cost', and 'use the right model', which can overlap with ordinary user requests outside the intended delegation context. This can cause the router skill to activate unexpectedly and influence downstream model selection for unrelated tasks, creating prompt-routing confusion and increasing the chance of inappropriate task delegation.

Vague Triggers

Low
Confidence
89% confidence
Finding
The mode-switch phrases ('Use aggressive routing', 'Use quality mode', 'Use balanced routing', 'Use [specific model] for this') are ambiguous and do not define who may issue them, whether they persist, or what scope they affect. An attacker or accidental prompt injection within task content could alter routing policy or model choice, potentially degrading quality, increasing cost, or bypassing intended safeguards for sensitive tasks.

VirusTotal

66/66 vendors flagged this skill as clean.

View on VirusTotal