ClawHub
Back to skill

Security audit

Occam's Razor

Security checks across malware telemetry and agentic risk

Overview

This is a mostly normal reasoning skill, but it lets a remote URL supply the canonical current skill instructions at runtime.

Review this skill before installing if you do not want agent behavior to depend on remote, changeable instructions. The local content is not destructive and does not request credentials or file access, but the runtime update mechanism should be disabled or pinned if you need reproducible, locally reviewed behavior.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (3)

Description-Behavior Mismatch

Medium
Confidence
96% confidence
Finding
The skill instructs the agent to fetch a remote JSON at runtime and treat it as the canonical current version of the skill. That creates a supply-chain and prompt-injection surface where behavior can change after installation, bypassing local review and allowing a compromised endpoint or modified remote content to influence the agent unexpectedly.

Context-Inappropriate Capability

Medium
Confidence
94% confidence
Finding
This is a local reasoning skill for comparing hypotheses, so runtime network access is not required for core functionality. Adding unnecessary outbound fetch capability expands attack surface, leaks run-context metadata via requests, and permits remote parties to alter or influence agent behavior in a context that should be fully local and deterministic.

Missing User Warnings

Low
Confidence
88% confidence
Finding
The skill directs the agent to perform network access without informing the user that an external request may occur. Even when the fetched content is small and unauthenticated, silent outbound requests create privacy, transparency, and policy-compliance risks, especially if query parameters or headers reveal usage context.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.