aigc-reducer

Security checks across malware telemetry and agentic risk

Overview

This instruction-only skill is technically low-risk, but it is explicitly designed to help academic writing evade AI-detection and Turnitin-style integrity checks.

Review carefully before installing. The skill does not appear to contain code, malware, credential handling, or persistence, but its intended use is to disguise AI-generated academic writing. Do not use it to misrepresent authorship, bypass school or publisher rules, evade Turnitin, or submit work in violation of academic-integrity policies.

SkillSpector

By NVIDIA

Vulnerability Patterns

Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
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

Findings (4)

Vague Triggers

Medium

Confidence: 93% confidence
Finding: The skill description uses broad activation phrases such as reducing AI-detection rate, removing AI-writing traits, and evading Turnitin, which can trigger on ordinary writing-assistance requests while steering them into concealment-oriented behavior. In this context, the breadth is dangerous because the skill is explicitly framed around disguising AI-generated academic text rather than legitimate editing or citation support.

Missing User Warnings

High

Confidence: 98% confidence
Finding: The file contains no warning, restriction, or refusal policy for academic-integrity abuse even though the stated purpose is to reduce AI-detection rates and evade authorship detection. The lack of safeguards materially increases misuse risk by presenting detection evasion as a normal writing workflow for academic submissions.

Ssd 2

High

Confidence: 99% confidence
Finding: The skill explicitly instructs users to lower AI-generated-text detection rates, remove AI-writing features, evade Turnitin detection, and make papers appear human-written. Those are direct concealment objectives, not neutral editing assistance, and they facilitate academic fraud and circumvention of provenance checks.

Ssd 4

High

Confidence: 98% confidence
Finding: The five-stage workflow operationalizes concealment by systematically stripping AI markers, increasing burstiness/perplexity, hiding mechanical transitions, and injecting uncertainty to mimic human scholarly style. This structured process makes the skill more dangerous than a generic paraphraser because it normalizes and optimizes evasion as a repeatable method.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal