Security audit

Humanizer.Local.Backup

Security checks for vulnerabilities and agentic risk

Overview

This text-editing skill is transparent about rewriting AI-sounding text, but it is purpose-built to remove AI-origin signals and lacks safeguards for contexts where authorship disclosure matters.

Review this carefully before installing. It does not contain malware-like behavior, but it is designed to make AI-generated text appear more human and may edit documents using file tools. Use it only where rewriting for style is appropriate, and do not use it to evade disclosure rules, academic integrity requirements, platform policies, or to invent a personal voice the author did not provide.

A.I.G

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

SkillSpector

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 (5)

Natural-Language Policy Violations

High

Confidence: 97% confidence
Finding: The core function explicitly aims to remove signs of AI-generated writing and make text sound human-written, but it includes no transparency, consent, or policy safeguards. That directly enables misrepresentation in settings like schoolwork, journalism, reviews, job applications, or compliance documentation, where the origin of text may be material.

Vague Triggers

Medium

Confidence: 94% confidence
Finding: The phrase "Or invoke directly when editing documents" does not clearly define what exact trigger, context, or constraints cause the skill to activate. Because it is broad and lacks negative examples or scoped conditions, it could overlap with many ordinary editing situations and lead to unintended invocation.

Vague Triggers

Medium

Confidence: 92% confidence
Finding: The skill description is broad enough to match ordinary editing or review requests, which increases the chance of unintended invocation. In this context, unintended use matters because the skill's purpose is to make text appear more human-written, which can facilitate undisclosed laundering of AI-generated content.

Missing User Warnings

Medium

Confidence: 95% confidence
Finding: The instructions aggressively encourage adding 'soul,' opinions, first-person perspective, and personality, yet provide no warning that doing so can fabricate personal stance or lived perspective not present in the source. In combination with the humanization goal, this increases the risk of deceptive rewriting that attributes human feelings or authenticity where none existed.

Vague Triggers

Medium

Confidence: 90% confidence
Finding: The task section says to humanize text whenever given text to humanize, but it does not define constraints, exclusions, or required user intent checks. That makes the skill easy to apply in routine editing flows without confirming whether the user is attempting to disguise AI-generated material or misrepresent authorship.

Static analysis

No suspicious patterns detected.