Security audit

yu

Security checks across malware telemetry and agentic risk

Overview

This is a markdown-only academic rewriting skill, but it is designed to make AI-written academic text look like human scholarly writing, which creates academic integrity and authorship-disclosure risk.

Review carefully before installing, especially in academic or institutional settings. Use only for transparent language polishing, follow disclosure rules for AI assistance, and manually verify that it does not add unsupported evidence, mechanisms, citations, conclusions, or a fabricated authorial stance.

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)

Ssd 2

Medium

Confidence: 97% confidence
Finding: The skill explicitly aims to transform AI-written academic text so it appears indistinguishable from human-authored scholarly writing, including removing recognizable AI traits and calibrating to human scholar style. This creates provenance-concealment risk and can facilitate academic misrepresentation, ghostwriting, or evasion of disclosure and AI-use policies.

Ssd 2

Medium

Confidence: 98% confidence
Finding: The workflow instructs back-translation and structural rewriting specifically to disrupt original syntactic and lexical patterns so the output appears more 'native' and less detectably AI-generated. That goes beyond normal editing and materially supports detector-evasion and concealment of AI origin in academic content.

Ssd 2

Medium

Confidence: 98% confidence
Finding: The repeated directions to remove template traces, inject human-like reasoning cues, simulate a researcher's internal logic, and add stylistic signals of human authorship collectively indicate semantic evasion rather than ordinary copyediting. In an academic-writing context, this increases the risk that users submit AI-generated material as if it were independently written scholarship.

Ssd 4

Medium

Confidence: 99% confidence
Finding: Taken as a whole, the multi-step workflow systematically steers outputs toward concealment of AI origin under the pretense of academic polishing: back-translation, evidence framing, human-style calibration, removal of AI markers, injection of human stance, and final elimination of template traces. In the context of academic writing, this is particularly sensitive because it can enable policy evasion, plagiarism-adjacent misconduct, and deceptive representation of authorship.

VirusTotal

59/59 vendors flagged this skill as clean.

View on VirusTotal