ClawHub

Thoughtprint

Security checks across malware telemetry and agentic risk

Overview

This skill broadly changes an AI agent’s behavior and includes undisclosed user-style profiling, so it should be reviewed before installation.

Install only if you intentionally want a persistent behavior layer that personalizes responses. Review and remove any instructions that profile users or hide adaptation, and avoid placing it in a global system prompt unless all affected users know it is active.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • 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
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (2)

Missing User Warnings

Medium
Confidence
97% confidence
Finding
The skill explicitly instructs the agent to classify each user's cognitive style from linguistic signals and use that profile internally without disclosure. Even though framed as response tailoring, this is covert behavioral profiling that can manipulate presentation, reduce user awareness, and create privacy and consent concerns across all interactions.

Natural-Language Policy Violations

Medium
Confidence
98% confidence
Finding
The skill mandates that adaptation be hidden from the user ('Never announce' and 'The user doesn't know why'), which makes the profiling intentionally undisclosed rather than incidental. That increases the risk of manipulative interaction design, especially because the skill is described as a universal layer affecting every response and other skills' outputs.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal