Back to skill

Security audit

Human Style Writing

Security checks for vulnerabilities and agentic risk

Overview

This is a prompt-only writing helper for human-style chats and social posts, with no code execution, persistence, credential access, or hidden data movement found.

Installers should understand that this skill is designed to make AI-written messages and posts sound more human. Use it transparently and avoid impersonation, fake experiences, or misleading social content; for formal legal, compliance, academic, business, or press writing, use a skill intended for that domain instead.

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
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • 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 (7)

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

These sections prescribe Chinese-language phrasing and style for specific platforms, but do not state that this applies only when the user wants Chinese output. Under the policy, forcing a specific language or locale without opt-in is a natural-language policy violation unless clearly documented and justified.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

Line L100 sets a default language preference of Chinese over English without offering a neutral language choice by default. This is a natural-language locale policy constraint that forces a specific language unless the user opts out.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

Line L109 instructs the skill to prefer Chinese for WeChat Moments without describing user choice or a documented, justified locale restriction. This creates a language policy issue because it imposes a locale default rather than letting the user opt in.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The off-scope guardrail does not actually preserve the scope boundary; it converts prohibited request types such as work email, academic writing, or legal/compliance language into allowed DM/social formats. This enables policy circumvention by laundering disallowed assistance through stylistic rewriting rather than refusing or maintaining the original restriction.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
89% confidence
Finding

The guidance says 'In CN, calibrate 您/你, 麻烦/请, and sign-off,' which introduces a locale-specific language policy in natural-language instructions. Because the file does not indicate that this is optional, user-selected, or limited to a region-specific use case, it can be read as enforcing a specific locale/register without opt-in.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
89% confidence
Finding

The checklist instructs reviewers to remove specific Chinese assistant phrases and calibrate Chinese pronoun/honorific usage, which imposes a locale-specific style expectation. Because the file does not state that it only applies to Chinese-language outputs or that language choice is user-driven, this can be read as forcing a particular language/locale convention without opt-in.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
50% confidence
Finding

SQP-3 applies to all file types and covers language or locale policy violations. This file is framed specifically as "CN/EN" guidance and focuses throughout on Chinese and English examples, which could be read as a locale constraint; however, it is a reference document rather than an explicit runtime instruction, so the concern is limited.

Content

No source excerpt is available for this finding.