ClawHub
Back to skill

Security audit

alter

Security checks across malware telemetry and agentic risk

Overview

This skill is a disclosed persona-building tool that handles very private user material, with privacy and persistence risks users should understand before use.

Install only if you are comfortable giving this skill personal writings, documents, chat exports, and possibly voice material for persistent persona memory. Review the actual companion services and installer before running them, check which model and voice providers are configured, and use the promised review/correction/delete controls for sensitive material.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (2)

Intent-Code Divergence

Medium
Confidence
93% confidence
Finding
The skill makes a strong privacy assurance that data stays on the local machine except limited outbound prompt traffic, but elsewhere authorizes use of an external configured build model for distillation and reconciliation. In a persona-building skill that ingests highly sensitive personal communications, this mismatch can mislead users into sharing intimate data under an inaccurate understanding of where processing occurs and what content may be transmitted off-device.

Ssd 3

Medium
Confidence
91% confidence
Finding
The skill explicitly treats every conversation as potential training data for a persistent persona. In context, this is especially sensitive because the skill is designed to model a user's identity, values, writing style, and private documents, so default ongoing reuse creates meaningful privacy, consent, and retention risks if users do not have granular control over what is learned or forgotten.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.