Skill flagged — suspicious patterns detected

ClawHub Security flagged this skill as suspicious. Review the scan results before using.

期刊论文AI率降低

v1.0.0

This skill should be used when the user wants to reduce AI detection rate in journal papers or academic documents. It transforms AI-generated content into hu...

2· 112·0 current·0 all-time
byAshi@innocencexy

Install

OpenClaw Prompt Flow

Install with OpenClaw

Best for remote or guided setup. Copy the exact prompt, then paste it into OpenClaw for innocencexy/anti-ai-skill.

Previewing Install & Setup.
Prompt PreviewInstall & Setup
Install the skill "期刊论文AI率降低" (innocencexy/anti-ai-skill) from ClawHub.
Skill page: https://clawhub.ai/innocencexy/anti-ai-skill
Keep the work scoped to this skill only.
After install, inspect the skill metadata and help me finish setup.
Use only the metadata you can verify from ClawHub; do not invent missing requirements.
Ask before making any broader environment changes.

Command Line

CLI Commands

Use the direct CLI path if you want to install manually and keep every step visible.

OpenClaw CLI

Bare skill slug

openclaw skills install anti-ai-skill

ClawHub CLI

Package manager switcher

npx clawhub@latest install anti-ai-skill
Security Scan
VirusTotalVirusTotal
Suspicious
View report →
OpenClawOpenClaw
Suspicious
medium confidence
!
Purpose & Capability
The name/description and the SKILL.md align: the skill's goal is to reduce AI-detection rates. However, it asks for adding 3–5 "真实论文原话引用" from recent core journals and tailoring outputs for indexed services (CSSCI, EI) without declaring any means to access those bibliographic databases or subscriptions. Requesting "real" citations and journal-sourced quotes is disproportionate to the declared environment (no credentials, no install) and suggests either an implicit expectation that the agent fabricate citations or that the user will supply sources—this mismatch is not documented.
!
Instruction Scope
The runtime instructions are tightly scoped to the stated purpose (rewriting, injecting typos/errors, adding citations, and targeting detectors). But they explicitly direct deceptive actions (evading detection, inserting deliberate fabrications or errors) and require measurements (AI-rate targets <15%) without specifying how to compute or verify AI-rate. The instructions also demand "真实的引用" but provide no sourcing, crawling, or verification steps—raising a risk the agent will invent citations or instruct the user to supply paid/credentialed content.
Install Mechanism
Instruction-only skill with no install spec, no code files executed, and no downloads—this is low-risk from an installation/execution perspective.
Credentials
The skill requests no environment variables or credentials, which is appropriate given there is no code. However, some declared objectives (integration with Turnitin/知网/维普, adding real citations from subscription databases) realistically require external access or credentials; the absence of any declared access or guidance is a proportionality mismatch and should be clarified.
Persistence & Privilege
No special persistence or elevated privileges requested (always:false, no config paths). The skill will not be forced into every agent run and does not declare autonomous elevation.
Scan Findings in Context
[no_scan_findings] expected: The static regex scanner returned no findings. That's expected because this is an instruction-only skill with no executable code for the scanner to analyze. Absence of findings does not imply the instructions are safe or coherent.
What to consider before installing
This skill is designed to help authors evade AI-detection systems and to alter academic content in ways that are deceptive (adding 'real' citations, inserting deliberate errors, and hitting AI-rate targets). Before installing or using: 1) Consider legal/ethical risks—this facilitates academic dishonesty and may violate publisher/institution policies. 2) Clarify how the skill will obtain and verify the "真实引用" and how it measures AI-rate (the SKILL.md gives targets but no verification method); if the skill expects you to provide subscription content or detector credentials, that should be explicit. 3) Beware of fabricated citations: if the skill cannot access authoritative databases, it may generate plausible-looking but false references. 4) If you still want the functionality, require the skill to (a) explicitly ask the user to supply legitimate source documents or credentials for any paid databases, (b) disclose how AI-rate is measured (which tool/endpoint), and (c) avoid fabricating citations. Given the deceptive nature and the documented mismatches, proceed only with caution and a clear understanding of the ethical and institutional consequences.

Like a lobster shell, security has layers — review code before you run it.

latestvk97dgbzwryp7adxa1b6y490vwh83p6sv
112downloads
2stars
1versions
Updated 1mo ago
v1.0.0
MIT-0

Anti-AI Skill — 期刊论文 AI 率降低(硕士化)

概述

本技能将 AI 生成的论文内容进行"人类化"处理,使其通过 AI 检测系统的同时具备硕士学位论文的学术水平。处理后的文本具有人类写作特征,包括适当的学术深度、真实的引用、刻意保留的微小瑕疵。


使用前提

用户需提供:

  • 待处理的 AI 生成论文内容(粘贴文本)
  • 目标期刊或检测系统(知网、维普、Turnitin 等)
  • 是否需要保留特定术语或表述

核心处理原则

1. 硕士学位学术语言

  • 使用规范的学术表述,避免过于"完美"的句式
  • 体现批判性思维:对既有研究进行评价与比较
  • 引入研究过程中的"试错痕迹"(如"最初尝试 X 方法,但因 Y 限制改用 Z 方案")
  • 适当使用领域"行话"和特定表述

2. 原文引用

  • 在文献综述或讨论部分添加 3-5 处真实论文原话引用
  • 引用格式符合目标期刊要求(APA、GB/T 7714 等)
  • 引用来源:近 5 年核心期刊、CSSCI、EI 收录论文
  • 引用时融入自己的分析和评述,而非简单罗列

3. 刻意保留的语言瑕疵

错误语序与修辞(≤10 处)

  • 状语位置略微偏移(如"通过对数据进行分析,我们得出结论" → "通过数据分析,我们结论得出")
  • 定语略微冗余或前置
  • 偶尔使用倒装句
  • 修辞手法轻微不当(比喻不太贴切、类比略牵强)

错别字(3-5 处,需标注): 在处理后的文本中故意保留以下类型错别字,并在输出中用括号标注正确写法:

错别字类型示例正确写法
的/地/得混淆"认真的学习""认真地学习"
的/地/得混淆"快速的完成""迅速地完成"
的/地/得混淆"研究的成果""研究的成果"(此句正确)
近形字"即今为止""迄今为止"
近形字"不可思意""不可思议"
同音字"反应问题""反映问题"
同音字"按排""安排"
漏字"中国中国""中国"
冗余"凯旋归来""凯旋"

标注方式:在错别字后用 [正确:X] 标注,例如:"通过研究 [正确:得] 出结论"


降 AI 率技术库

核心技巧(按效果排序)

1. 逻辑重构法 ⭐⭐⭐⭐⭐

  • 将"总-分-总"结构改为"问题-解决方案"结构
  • 调整段落顺序,将"影响"提到"背景"前论述
  • 用案例引入观点,代替平行论述

2. 同义词学术化替换 ⭐⭐⭐⭐⭐

  • 将 AI 常用词替换为学术表达:
AI 常用词学术化替换
值得注意的是从实验观察来看
首先、其次、最后第一、第二、第三(或另起炉灶)
此外另须指出
综上所述基于上述分析
非常重要具有关键意义
很多若干
很大提高显著提升

3. 句子结构重组 ⭐⭐⭐⭐⭐

  • 改变主被动语态
  • 拆分长句为短句
  • 合并短句为长句
  • 调整句子成分顺序

4. 注入个人观点与批判 ⭐⭐⭐⭐⭐

  • 加入对文献的评价与比较("Smith 的理论在 A 领域有效,但 Jones 指出其在 B 情境的局限性,笔者认为...")
  • 提出研究疑问("这一结论是否适用于跨文化背景?")
  • 关联实际案例或近期事件

5. 风格化表达 ⭐⭐⭐⭐

  • 交错使用长句(复杂逻辑)和短句(强调观点)
  • 适当使用不太贴切的比喻/类比
  • 在非严格数据部分使用少量修辞

6. 试错痕迹 ⭐⭐⭐⭐

  • 描述研究过程中的挫折与调整
  • 例如:"最初采用 X 方法进行实验,但因 Y 因素导致数据不稳定,遂改用 Z 方法"

7. 衔接词多样化 ⭐⭐⭐⭐

避免使用固定衔接词组合,多样化表达:

  • 转折:然而、不过、但同时
  • 递进:更进一步、在此基础上
  • 因果:因而、故而、由此可见
  • 总结:总之、总而言之

处理流程

Step 1:分析原文

  • 识别 AI 生成痕迹明显的段落
  • 统计各段落 AI 率(按轻度/中度/重度分类)
  • 标记需要保留的核心观点

Step 2:分段处理

对不同 AI 痕迹程度的内容采用不同策略:

AI 率处理策略
<30%人工微调 + 衔接词替换
30%-70%逻辑重构 + 同义词替换 + 结构重组
>70%深度改写 + 文献融合 + 个人观点注入

Step 3:应用降 AI 技术

按顺序执行:

  1. 逻辑重构
  2. 同义词替换
  3. 句子结构重组
  4. 注入批判性思维
  5. 添加原文引用
  6. 植入错别字(3-5 处)
  7. 植入语序/修辞错误(≤10 处)

Step 4:质量验证

  • 整体 AI 率目标:<15%
  • 局部 AI 率目标:<25%
  • 人工通读:检查语义连贯性
  • 格式检查:引用格式、术语统一

输出格式

处理后的文本

【处理后的论文内容】

---

【错别字清单】(共 X 处)
1. 第 X 段:"XXX" `[正确:X]`
2. 第 X 段:"XXX" `[正确:X]`
...

【语序/修辞错误清单】(共 X 处)
1. 第 X 段:XXX(原因说明)
2. ...

【新增引用】
1. [作者, 年份]: 具体内容...
2. ...

【技术使用记录】
- 逻辑重构:X 处
- 同义词替换:X 处
- 句子结构重组:X 处
- 注入个人观点:X 处
- 新增引用:X 处
- 植入错别字:X 处
- 植入语序错误:X 处

注意事项

  1. 学术准确性:所有改写必须保留核心学术观点的准确性
  2. 逻辑通顺:即使植入错误,也要保证整体逻辑可读
  3. 适度原则:错别字和错误不宜过多,以免影响阅读体验
  4. 格式统一:引用格式、术语使用保持全文一致
  5. 检测目标:根据目标检测系统(知网/维普/Turnitin)调整策略

参考资源

文件用途
references/ai_reduction_techniques.md降 AI 率技术详细说明与示例

Comments

Loading comments...