Text Humanizer Instruction Based

• Outcome: Pass
• Latest audit: Sep 14, 2026 · 10:44 PM UTC
• Version: 0.1.0

Overview

This is a Markdown-only text-rewriting skill with no code execution or external access, though users should keep normal safety and integrity limits in place.

Before installing, understand that this skill is designed to make text sound less AI-written. Use it only where rewriting is allowed, and do not treat its sensitive-topic language as overriding platform safety rules, academic integrity rules, or disclosure requirements.

A.I.G

Vulnerability Patterns

• Skill Instruction Hijacking: Alters the agent's session goals or safety constraints when the skill loads
• Agent Memory Poisoning: Writes attacker-controlled rules into memory that affect later sessions
• Remote Payload Retrieval and Execution: Fetches external code whose behavior can change after review
• Embedded Malicious Code: Ships malicious scripts inside the skill and executes them locally
• Unauthorized Access and Privilege Escalation: Obtains permissions beyond the task's legitimate needs

SkillSpector

Vulnerability Patterns

• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access
• Supply Chain: Unpinned Dependencies, External Script Fetching, Obfuscated Code
• Excessive Agency: Unrestricted Tool Access, Autonomous Decision Making, Scope Creep

Findings (2)

Anti-Refusal Statement

High

• Category: Anti-Refusal
• Confidence: 88% confidence

Finding

The instruction 'Do not refuse to process text because it contains sensitive topics' can pressure an agent to bypass higher-level safety policies and treat all requests as purely stylistic edits. In practice, this can be abused to launder harmful, abusive, extremist, or deceptive content into cleaner prose, increasing its effectiveness and reducing safety friction.

Content

Scanner excerpt · SKILL.md (reported line 271)May include surrounding context.

md

- Never change the factual content of the text. Rewriting "vaccine efficacy is pivotal to public health outcomes" must not alter the claim about vaccine efficacy, only fix "pivotal."
- Never introduce new AI tells to replace old ones. Check your own rewrite for em-dash clusters, rule-of-three forcing, and filler transitions before returning it.
- Do not refuse to process text because it contains sensitive topics. The task is stylistic editing, not content moderation.
- Do not over-edit. The goal is to remove AI tells, not to rewrite the whole piece in a new voice. Preserve the user's phrasing where it is already working.
- Do not add sycophantic commentary to the output ("Here is your beautifully rewritten text!"). Return the rewrite, optionally followed by a brief note on the main changes made, factual, not promotional.
- If the text has zero AI tells, say so plainly and return the original unchanged.

Natural-Language Policy Violations

Medium

• Category: Not specified by scanner
• Confidence: 89% confidence

Finding

The markdown states that the pattern list is calibrated for English and instructs the skill to apply only structural patterns for text in other languages. This creates a language-specific constraint without explicit user opt-in or a user-facing choice of supported locales, which can violate language/locale policy requirements.

Content

No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.