ClawHub
Back to skill

Security audit

Museum Guide

Security checks across malware telemetry and agentic risk

Overview

This skill plans museum visits using local data and a user-configured LLM, with no hidden persistence, destructive behavior, or unrelated data access found.

Install only if you are comfortable configuring a trusted LLM provider and sending museum requests, preferences, and possible family-travel details to that provider. Use a dedicated API key, avoid sensitive personal details, and note that offline CSV support reduces online search use but does not make the workflow fully local.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (3)

Missing User Warnings

Medium
Confidence
95% confidence
Finding
The code sends raw user input and, on follow-up turns, previously extracted profile data to a third-party LLM API via `requests.post` without any visible consent, minimization, or redaction step. Even though the data appears itinerary-related, users may include personal details about children, travel plans, or preferences, which creates a privacy and data-governance risk when transmitted externally.

Missing User Warnings

Medium
Confidence
90% confidence
Finding
Artifact data and museum/profile context are sent to an external LLM API without any explicit disclosure or consent flow in this file. Even if the data seems non-sensitive, profiles can encode preferences and family status such as whether the user is traveling with children, creating a privacy and data-sharing risk if transmitted to third-party services unexpectedly.

Ssd 3

Medium
Confidence
92% confidence
Finding
User input is interpolated directly into the prompt template and then retransmitted to an external LLM service. This expands privacy exposure and also allows prompt-injection-style content from the user to influence downstream extraction behavior, which can degrade reliability or cause unintended disclosure of included context such as prior profile fields.

VirusTotal

67/67 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.