by xiaofenggan01 · Claude Skill · ★ 410
Last updated: · Indexed by AgentSkillsHub · Auto-synced every 8h
AIGC Reduce 降低学术论文 AIGC 查重率的 Claude Code Skill。 基于知网 3.0、万方、PaperPass、PaperPure 等主流检测器的技术原理,结合 humanizer skill(源自 Wikipedia "Signs of AI writing")的 AI 痕迹识别方法论,通过三轮降重协议(去除痕迹 → 注入人类特征 → 自检审计)让文本呈现「人类写作」的统计特征。 安装 安装后,当你在 Claude Code 中提到「降 AI 味」「降 AIGC 率」「去 AI 检测」等关键词时,该 skill 会自动触发。 为什么需要这个工具? AIGC 检测器已经在各大高校强制使用。但检测技术存在根本性局限: 写得好 = 判 AI? 高质量流畅论文假阳性率 6.5%,有口语化瑕疵的论文仅 1.8% 跨平台结果不一致: 同一篇论文知网 AI 率 35%,万方仅 12% 术语密集型论文受害: 术语密度 20% 时假阳性率 8-12% 检测器依赖浅层特征: DPO 训练可使检测准确率下降 60% 核心思路:不是帮人作弊,而是帮本来是人写的论文降低被误判的风险。 三条铁律 禁止 AI 全量重写 — AI 重写 AI 文本 = 叠加 AI 指纹(PaperPure 实测:重写后检测率升至 100%) 修改率必须 40% — 轻度修改(40%)可达 60-80% 确定性替换 — 每次只改一小处,不经过 LLM 的 token 采样 三轮降重协议 第一轮:去除 AI 痕迹(减法) 扫描文本 → 词级替换 → 句级重构 → 段落调整,确保修改率 40%。 替换操作覆盖三个层级: 词级(10-15%):模板词、72 个中文 AI 高频词、动词的系统性替换 句级(15-20%):语序重组、长句拆分、被动改主动、主语更换 段落级(10-15%):对称段长打破、段间过渡、语义注入 第二轮:注入人类特征(加法) 源自 humanizer skill 的核心洞察:只去除 AI 痕迹不够,干净但无灵魂的文本和 AI 文本一样容易被检测。 节奏工程:故意制造长短句交错,目标突发性 CV 0.45 不确定性注入:"推测…""猜测…""不过还需要更多数据支持" 操作细节补充:加入实验观察、参数波动、异常说明 声音校准:如果用户提供参考文本,分析并匹配其写作风格 第三轮:Anti-AI 审计(自检) 审视文本,回答"这段话为什么还像 AI 写的?",逐项排查 10 种深度 AI 痕迹模式: 重要性膨胀("至关重要""不可忽视") 同义词轮换(同一概念被叫了 3 个以上名字) 三板斧强迫症(连续出现"A、B和C"三件套) 系词回避(AI 不用"是",偏用"作为/体现/代表") 模糊归因("有研究表明"无引用)
| Stars | 410 |
| Forks | 18 |
| Language | Python |
| Category | Claude Skill |
| License | MIT |
| Quality Score | 63.2831373484888/100 |
| Open Issues | 1 |
| Last Updated | 2026-06-28 |
| Created | 2026-05-02 |
| Platforms | claude-code, python |
| Est. Tokens | ~6k |
Looking for a aigc-reduce alternative? If you're comparing aigc-reduce with other claude skill tools, these 6 projects are the closest alternatives on Agent Skills Hub — ranked by topic overlap, star count, and community traction.
How real engineers run Claude Code and Codex: spec-driven planning, enforced TDD, persistent memory, and quali
My personal Claude Code and OpenAI Codex setup with battle-tested skills, plugins, hooks and agents that I use
Repo for the Claude Code Marketplace to use with the Claude for Life Sciences Launch. This will continue to ho
Report issues related to MCP integration with Claude here.
A tool for autonomous agents like OpenClaw to discover and learn skills autonomously
Open-source AI marketing plugin for agencies & in-house teams — 158 skills, 25 specialist agents, 12-Part Stra
Explore other popular claude skill tools:
aigc-reduce is 降低学术论文 AIGC 查重率的 Claude Code Skill | A Claude Code skill for reducing AIGC detection rates in academic papers. It is categorized as a Claude Skill with 410 GitHub stars.
aigc-reduce is primarily written in Python.
You can find installation instructions and usage details in the aigc-reduce GitHub repository at github.com/xiaofenggan01/aigc-reduce. The project has 410 stars and 18 forks, indicating an active community.
aigc-reduce is released under the MIT license, making it free to use and modify according to the license terms.
The top alternatives to aigc-reduce on Agent Skills Hub include pilot-shell, claude-codex-settings, life-sciences. Each offers a different approach to the same problem space — compare them side-by-side by stars, quality score, and community activity.