ClawHub
Back to skill

Security audit

Systems Thinking

Security checks across malware telemetry and agentic risk

Overview

This is a disclosed systems-thinking prompt skill with a user-run installer that links it into known agent skill directories.

Before installing, note that running the installer may link this skill into every detected supported agent, and --force can replace an existing same-named install. Use the installer flags to restrict the target agents or location, and treat the disturbance retry example as conceptual rather than copy-ready until corrected.

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 (2)

Intent-Code Divergence

Medium
Confidence
94% confidence
Finding
This is a real issue because the prose claims a measurable health-check-based feedforward control path, but the code only flips a local `_healthy` flag after request outcomes and never actually queries an independent health signal. That mismatch can mislead downstream users into believing the example demonstrates proactive disturbance compensation when it actually implements a simplistic local circuit-breaker heuristic, which may cause incorrect designs and unsafe retry behavior under failure conditions.

Intent-Code Divergence

Low
Confidence
89% confidence
Finding
This is a true defect because the sample references `attempt` in the exception path without defining it in that scope, so the demonstrated backoff logic is not actually valid code. In a reasoning skill that teaches control patterns for production retries, broken sample code can propagate flawed implementations, disable damping logic entirely, or cause developers to copy a pattern that fails at runtime.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.