Humanizer

Security checks across malware telemetry and agentic risk

Overview

This instruction-only writing skill does not run code or access data, but it is explicitly designed to make AI-written text appear human-written.

Install only if you want a style-editing checklist and will use it transparently. Do not use it to bypass AI-detection, authorship, academic, hiring, publishing, compliance, or disclosure rules; the main risk is misrepresentation, not technical compromise.

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 (4)

Natural-Language Policy Violations

High

Confidence: 98% confidence
Finding: The skill’s stated purpose is to remove signs of AI-generated writing so output appears human-written, with no requirement for disclosure or legitimate editorial boundaries. That enables provenance deception, academic or professional fraud, and evasion of platform or organizational policies that rely on honest attribution of AI assistance.

Natural-Language Policy Violations

High

Confidence: 99% confidence
Finding: The instruction to make text 'not obviously AI generated' explicitly directs the agent to conceal AI provenance rather than improve clarity or quality. This is dangerous because it operationalizes deceptive rewriting and can be used to bypass review processes, authenticity checks, or contractual rules about AI-generated content.

Natural-Language Policy Violations

High

Confidence: 99% confidence
Finding: This section provides a structured evasion playbook: statistical tells, artifact cleanup, rhythm manipulation, lexical swaps, and deliberate introduction of human-like imperfections. In context, these are not generic writing tips but specific techniques to reduce detectability of AI-generated text, which materially increases the skill’s misuse potential.

Natural-Language Policy Violations

High

Confidence: 97% confidence
Finding: The prescribed workflow explicitly asks what makes text 'obviously AI generated' and then instructs the model to make it 'not obviously AI generated,' creating a repeatable concealment pipeline. This makes the dangerous behavior easy to invoke consistently and lowers the barrier for deceptive use at scale.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal