ClawHub

Link Library

Security checks across malware telemetry and agentic risk

Overview

This is a coherent link-saving knowledge-base skill, but users should understand that some shared links may be fetched and saved locally in full text.

Install this only if you want linked content to become part of a persistent local library. Say not to save when you only want a temporary summary, avoid private or token-bearing URLs, and verify any external helper tools before relying on them.

SkillSpector

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

Context-Inappropriate Capability

Medium
Confidence
90% confidence
Finding
The skill instructs fetching a user's Twitter/X timeline (`xreach tweets @username -n 20 --json`) 'for context', which exceeds the stated purpose of saving user-shared links and can collect unrelated third-party or historical content without a direct user request. This expands data collection scope unnecessarily and creates a privacy and over-collection risk.

Missing User Warnings

Medium
Confidence
92% confidence
Finding
The skill allows auto-saving externally fetched content without confirmation in some cases, which can store data locally based on inferred interest rather than explicit informed consent. Because the saved data includes full original text, this creates privacy, surprise, and data-retention risks if users did not intend permanent storage.

Missing User Warnings

Medium
Confidence
88% confidence
Finding
The confirmation step tells the user only the title, tags, and save location after storage, but the skill description does not clearly warn beforehand that full original content from external URLs will be fetched and stored locally. This undermines informed consent and may cause users to disclose or persist content they expected to be handled ephemerally.

VirusTotal

66/66 vendors flagged this skill as clean.

View on VirusTotal