ClawHub

Algorithmic Art

Security checks across malware telemetry and agentic risk

Overview

This is a generative-art helper with some branding and CDN caveats, but no hidden data access, persistence, or destructive behavior.

Install only if you are comfortable with a skill that strongly preserves an Anthropic-like UI template and whose generated HTML may contact CDNs for p5.js and fonts. For private or offline use, ask the agent to remove third-party branding and bundle or replace external resources.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • 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
Findings (4)

Context-Inappropriate Capability

Medium
Confidence
93% confidence
Finding
The skill hard-codes preservation of Anthropic branding and exact template styling even though the stated purpose is generating original algorithmic art. This creates instruction-level policy drift: the agent is being steered to reproduce unrelated proprietary presentation elements, which can cause unauthorized brand usage, reduce user control, and encourage copying beyond what is necessary for the art task.

Context-Inappropriate Capability

Medium
Confidence
91% confidence
Finding
Requiring the agent to use an existing template as the literal starting point conflicts with the skill's originality objective and can push the model to reproduce substantial nonfunctional template content verbatim. That increases the risk of unauthorized copying of template structure and presentation, especially when the copied material is not essential to fulfilling the user's art request.

Description-Behavior Mismatch

Low
Confidence
95% confidence
Finding
The template claims the artifact should be self-contained, but it loads JavaScript and fonts from external CDNs. This introduces a supply-chain and privacy risk: if the CDN content is changed, unavailable, or blocked, the viewer can execute untrusted code or fail to render, and requests to third parties may leak user metadata.

Natural-Language Policy Violations

Medium
Confidence
90% confidence
Finding
Forcing Anthropic branding and styling without user choice or task necessity can misrepresent origin, create unauthorized affiliation signals, and override user intent. In this context, the danger is amplified because the requirement is framed as mandatory and exact, which makes the agent prioritize brand replication over the actual generative-art task.

VirusTotal

59/59 vendors flagged this skill as clean.

View on VirusTotal