ClawHub
Back to skill

Security audit

PBE Extractor

Security checks across malware telemetry and agentic risk

Overview

This is an instruction-only text analysis skill with one file-writing wording inconsistency, but no code or install behavior that would create files automatically.

Safe for normal text-analysis use. Avoid pasting confidential content unless your agent and model provider are approved for it, and do not allow the agent to create a review file unless you explicitly want that local artifact.

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
90% confidence
Finding
The skill states that it does not write files to disk, but later instructs that drifted normalization results be added to `requires_review.md`. This creates a contradictory data-handling contract that could cause an agent runtime with file-write capabilities to persist user-derived content unexpectedly, violating user expectations and potentially leaking sensitive input to local storage.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.