Phy Content Humanizer Audit

• Outcome: Review
• Latest audit: Sep 12, 2026 · 7:58 AM UTC
• Version: 1.0.3

Overview

This skill is locally implemented and disclosed, but it is designed to help users reduce platform AI-detection signals before publishing social posts.

Review before installing if you publish on platforms with AI-content or authenticity policies. The code appears local and limited to text analysis, but its main value is helping content look less AI-detectable, so avoid using it to mislead audiences or bypass platform rules.

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

• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• MCP Least Privilege: Underdeclared Capability, Wildcard Permission, Missing Permission Declaration
• 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

Findings (7)

Natural-Language Policy Violations

High

• Confidence: 98% confidence
• Finding: The file header explicitly frames the tool as measuring signals platforms use to detect AI-generated content and helping users identify which signals to change. That makes the skill an optimization aid for evading platform integrity and moderation systems, even if it does not directly rewrite text itself.

Natural-Language Policy Violations

High

• Confidence: 99% confidence
• Finding: The recommendations give concrete, actionable steps for making text appear more human and lowering an 'AI signature,' such as adding contractions, pronouns, questions, and removing flagged words. This is effectively coaching for stealth/evasion against social-platform AI-content detection systems.

Lp3

Medium

• Category: MCP Least Privilege
• Confidence: 84% confidence
• Finding: The skill advertises command examples that read local files (`--file draft.txt`) and piped stdin, but it does not declare any tool scope or permissions boundary. In an agent ecosystem, undeclared file-read capability weakens reviewability and can lead to broader-than-expected access to local content when the skill is invoked.

Ssd 2

Medium

• Confidence: 97% confidence
• Finding: Although framed as an 'auditor,' the skill explicitly tells users which platform-detection signals to change so content is less likely to be flagged as AI-generated. That is effectively evasion guidance for trust-and-safety or ranking systems, enabling users to disguise synthetic content rather than merely assess it.

Ssd 2

Medium

• Confidence: 99% confidence
• Finding: This section provides step-by-step remediation actions—add contractions, add specific data, remove flagged words—and explicitly states these changes lower the 'AI signature' by 20-30%. That is concrete, actionable bypass guidance for stylometric AI-detection systems, which increases the likelihood of deceptive or policy-evading use.

Ssd 4

Medium

• Confidence: 96% confidence
• Finding: The introductory text presents the tool as an auditor for the exact signals used by platforms to detect AI-generated content. In context, that framing supports iterative testing and refinement of content until it passes, which lowers the effectiveness of authenticity controls.

Ssd 4

Medium

• Confidence: 99% confidence
• Finding: The top-fix recommendations are explicit coaching instructions for reducing detection risk by altering linguistic markers associated with AI text. That materially increases the usefulness of the tool for incremental evasion, especially when combined with per-platform verdicts and thresholds.

Static analysis

No suspicious patterns detected.