ClawHub
Back to skill

Security audit

Remove the AI flavor

Security checks across malware telemetry and agentic risk

Overview

This writing skill is not malware, but it explicitly teaches users how to make AI text pass as human-written and pass AI-detection checks.

Review carefully before installing. Use this only for transparent style editing of text you are allowed to revise; do not use it to misrepresent authorship, bypass academic or platform rules, or fabricate research support. Avoid uploading sensitive drafts to third-party detector or rewriting services.

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 (4)

Context-Inappropriate Capability

High
Confidence
98% confidence
Finding
This section explicitly provides a workflow to reduce AI-detection scores, including detector verification and iteration until threshold targets are met. That goes beyond style editing and becomes operational guidance for evading provenance checks, which can facilitate academic dishonesty, spam, and deceptive content laundering.

Context-Inappropriate Capability

High
Confidence
99% confidence
Finding
The skill suggests adding hypothetical data or cases so text appears based on real research. That encourages fabrication of supporting evidence and can be used to make false or deceptive content look credible, increasing risks of misinformation, fraud, and academic misconduct.

Ssd 2

Medium
Confidence
97% confidence
Finding
These sections contain prompts and procedures specifically aimed at making AI-generated text appear human-written and bypass AI-origin heuristics. In context, the skill is not merely improving readability; it is packaging concealment tactics that can be used to misrepresent machine-generated work as authentic human authorship.

Ssd 4

High
Confidence
99% confidence
Finding
The multi-step workflow operationalizes evasion by combining rewriting, external de-AI tools, detector testing, threshold goals, and iterative tuning until the content passes. This is especially dangerous because it lowers the barrier for systematic deception and enables repeatable laundering of AI-generated or otherwise non-original content.

VirusTotal

66/66 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.