Security audit

Humanizer

Security checks for vulnerabilities and agentic risk

Overview

This skill is a simple writing editor, but its stated purpose is to hide signs of AI-generated writing without disclosure guardrails.

Review before installing if you work in contexts where AI assistance must be disclosed. This skill has no code or persistence, but its editing goal is to make AI-written text look human-written, so use it only for legitimate style improvement and not to bypass authorship, academic, workplace, legal, or platform disclosure rules.

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: 98% confidence
Finding: The core function is to remove signs of AI-generated writing and make text appear human-written, which directly enables concealment of AI involvement. That can facilitate deception in academic, professional, journalistic, compliance, or trust-based settings where readers rely on honest disclosure of authorship.

Vague Triggers

Medium

Confidence: 93% confidence
Finding: The phrase "Or invoke directly when editing documents" does not specify what exact trigger phrases, contexts, or exclusions cause the skill to activate. Because it is broad and lacks constraints or negative examples, it could lead to unintended invocation during ordinary document editing.

Vague Triggers

Medium

Confidence: 95% confidence
Finding: The skill description advertises a broad capability to make text sound more natural and human-written without defining safe-use boundaries. That ambiguity increases the chance the agent will activate in disclosure-sensitive contexts, including polishing AI output so it can be misrepresented as authored by a human.

Missing User Warnings

Medium

Confidence: 94% confidence
Finding: The skill contains extensive instructions for making text feel more human and personal, but it does not warn users against deceptive use or mention disclosure-sensitive contexts. This omission makes misuse more likely because the guidance actively improves plausibility while providing no ethical or policy guardrails.

Vague Triggers

Medium

Confidence: 88% confidence
Finding: The task instruction "When given text to humanize" is an extremely broad trigger with no constraints on when the skill should refuse or seek clarification. In practice, this can cause the agent to apply deceptive rewriting to any supplied content, including regulated, academic, or identity-sensitive communications.

Static analysis

No suspicious patterns detected.