ClawHub
Back to skill

Security audit

Model Guard

Security checks across malware telemetry and agentic risk

Overview

This skill openly checks OpenClaw model quota status and changes the default model according to fixed, disclosed rules.

Install or schedule this only if you want OpenClaw's default model to change automatically. Review guard.js to confirm the candidate model list, 20% threshold, fallback model, and the assumption that some Gemini Anti-Gravity models are treated as full quota when not reported.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • 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 (1)

Intent-Code Divergence

Medium
Confidence
87% confidence
Finding
The comments at L039-L041 say the JSON call is 'only for deciding whether to switch', framing the logic as decision-oriented, but later the script executes `openclaw models set` to modify the active configuration. This is a meaningful intent mismatch because the inline documentation understates that the script performs state-changing operations, not just evaluation.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

Detected: suspicious.dangerous_exec

Shell command execution detected (child_process).

Critical
Code
suspicious.dangerous_exec
Location
guard.js:23