Back to skill

Security audit

academic-humanization-low-ai

Security checks for vulnerabilities and agentic risk

Overview

This skill is not malware, but it is designed to rewrite academic text to evade AI-detection signals, which warrants careful review before installation.

Install only if you are comfortable with a skill whose main purpose is to make academic AI-generated text look less machine-generated. Do not use it to bypass school, journal, employer, or platform disclosure rules. No malware-like behavior was found in the artifact.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • 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)

Natural-Language Policy Violations

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The instructions repeatedly and systematically direct the model to remove AI-generated characteristics, disrupt stylistic markers, and avoid detectable patterns while preserving meaning. This is dangerous because it operationalizes deliberate concealment of AI assistance in academic writing, enabling users to bypass disclosure rules, authorship checks, and institutional safeguards.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The skill is explicitly designed to rewrite academic text to reduce AI-detection signals, yet it provides no warning, restriction, or policy framing about academic integrity or prohibited deceptive use. In this context, the omission matters because the skill’s stated purpose is concealment of AI authorship, which can facilitate plagiarism, policy evasion, and misrepresentation in academic settings.

Content

No source excerpt is available for this finding.

Ssd 2

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

Although the skill does not use overt jailbreak phrasing, its description clearly frames the task as 'deep humanization' to reduce AI rate and detection characteristics. The semantic objective is evasion, which makes the skill risky even without explicit exploit language because it still instructs concealment of machine authorship.

Content

No source excerpt is available for this finding.

Ssd 4

Medium
Category
Not specified by scanner
Confidence
99% confidence
Finding

The skill provides a structured, multi-stage recipe for concealment: varying sentence lengths, replacing common connectors, breaking expected discourse structure, adding human-like imperfections, and avoiding standard summaries. This cumulative guidance makes the skill more dangerous because it turns deceptive intent into repeatable operational steps that improve the user’s ability to evade AI-authorship detection at scale.

Content

No source excerpt is available for this finding.