qzkinhit

Research Builder — AI skill for Claude Code

AI community

Research and paper-writing skill for Claude Code and Codex: faithful baselines, rigorous experiments, bilingual LaTeX papers, de-AI writing, figures and tables.

How to install Research Builder

This entry records only its repository, not the path inside it, so there is no exact command to give. Open qzkinhit/research-builder and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Research Builder does

Research and paper-writing skill for Claude Code and Codex: faithful baselines, rigorous experiments, bilingual LaTeX papers, de-AI writing, figures and tables.

Alternatives in AI

  • Welcome — AI Research Skills — You now have access to 86 production-ready skills covering the entire AI research lifecycle: literature survey 5.4k ★
  • Research Paper Writing Skills — Skill package for ML/CV/NLP paper writing, curated and adapted from Prof 1.5k ★
  • Academic Humanizer — Strip AI-writing tells from papers and grant proposals (NSF/NIH), while keeping scholarly voice and tying clai 1.1k ★

README

research-builder

research-builder 是一套给 Claude Code 与 Codex 用的科研辅助 skill。它把一位研究者在数据质量与数据准备方向写顶会与汇刊论文时形成的五阶段工作流程、写作纪律和检查工具整理成规则,让 AI 在研究者的主导下承担其中的工程实现、实验记录、排版、润色和核查。

研究由你主导。选题、文献调研与精读、核心想法、baseline 的取舍、方法改动和论文里的每一个论断,都由你提出或确认。skill 按五个阶段推进,每个阶段结束时交出可核验的产物,等你确认后再进入下一阶段。所有数字只来自真实的实验记录,它不会也不应该替你编造结果。

它适用于「提出方法,用理论与实验证明它更好」的实证型研究。你放进来的素材越多(你读过的文献、你的论文、你的图表与文字),它的产出越贴近你的研究与风格。

*English summary.* research-builder is a Claude Code and Codex skill set that assists a researcher through a five-stage workflow for empirical papers. The researcher chooses the topic, reads the literature, proposes the idea and approves every stage. The skill helps turn the researcher's surveyed related work into a runnable system, faithfully reproduced baselines, multi-seed experiments and a bilingual (English and Chinese) LaTeX paper whose numbers are filled in by a backfill script from the experiment records. It ships a writing manual with 32 AI-writing patterns adapted from [humanizer](https://github.com/blader/humanizer) for academic prose, checkers for LaTeX, Word and PowerPoint, a matplotlib module for print-size experiment figures, LaTeX table macros, a bilingual paper skeleton, rebuttal and grant-proposal guides, and an editable-PPT figure skill with a starter SVG icon set. The documents are written in Chinese.

能做什么

能力 内容
科研五阶段 对齐 → 搭系统(纯方法与应用薄壳分离) → 忠实复现 baseline(三层忠实度认证) → 实验(公认数据集、等预算、多种子) → 诊断改方法 → 回填论文
写作与去 AI 腔 叙事与取舍手册,以及改编自 humanizer 的 32 条 AI 腔模式、改写流程和声音匹配。检查脚本覆盖 LaTeX、Word、PowerPoint 和 Markdown
论文骨架 方法论文从标题到附录的逐节写法和中英文句式,benchmark 论文的叙事,基金本子与 rebuttal 的写法
实验图与表 按印刷尺寸出图的绘图模块和四个样例,能力表、数据集表和主表的宏与生成脚本
学术插图 独立的 paper-figure-pptx 技能,做可编辑的动机图、系统图与本子图,带 16 个自绘 SVG 入门图标
模板 中英镜像的 LaTeX 骨架、镜像核对脚本、数字回填脚本、rebuttal 模板
调研 独立的 paper-survey 技能,优先抓 arXiv 的 LaTeX 源与 HTML,按权威与开源筛选

快速开始

**第一步,安装。** 克隆到 Claude Code 的 skill 目录并运行安装脚本,它会把 `paper-survey` 和 `paper-figure-pptx` 也部署成独立 skill。

git clone https://github.com/qz