Claude代码技能，使用并行子智能写作40-100+页学术报告（FYP、论文、学位论文）——比顺序写作快3-4倍。

Research repo 在→编纂了 LaTeX 论文。86页，耗时~6小时，已验证。

它的作用

把它指向你的研究仓库。它启动并行代理，提取数据，同时写章节，编译LaTeX，并审计交叉引用：

比顺序写作快3-4倍。在一份86页的FYP报告中验证，耗时约6小时。

安装

作为克劳德代码技能

依赖关系

用途

用 Claude Code 中的 /write-report 调用，或描述你的报告需求，技能将被建议。

快速开始

1. 准备好带有代码+结果的研究资料库
2. 告诉克劳德：“基于这个仓库写我的FYP报告”
3. Claude 将自动执行三波流水线

引用验证

提交前将所有引用与3个学术数据库核实。捕捉AI幻觉引用（6-55%的AI生成引用为伪造）：

级联流水线：CrossRef（1.4亿+DOI）->语义学者（2亿+篇论文）->OpenAlex（2.4亿+作品）。

write-academic-report

Claude Code skill that writes 40-100+ page academic reports (FYP, thesis, dissertation) using parallel subagents — 3-4x faster than sequential writing.

Research repo in → compiled LaTeX thesis out. 86 pages in ~6 hours, validated.

What it does

Point it at your research repo. It launches parallel agents to extract data, write chapters simultaneously, compile LaTeX, and audit cross-references:

Wave 0: Data Preparation    Wave 1: Chapter Writing    Wave 2: Assembly
(5-6 parallel agents)       (3-4 parallel agents)      (sequential)

Codebase analysis            Ch1-2 (Intro + Lit Review)  Merge chapters
Experiment data              Ch3 (Methodology)           Cross-ref audit
Statistics                   Ch4-5 (Setup + Results)     Compile (tectonic)
Figure generation            Ch6 (Conclusion)            Quality review

3-4x faster than sequential writing. Validated on an 86-page FYP report produced in ~6 hours.

Install

As a Claude Code skill

# Copy to your skills directory
cp -r . ~/.claude/skills/write-report/

# Or symlink
ln -s $(pwd) ~/.claude/skills/write-report

Dependencies

# LaTeX compiler (required)
brew install tectonic    # macOS
# Or: cargo install tectonic

# Python 3.10+ for scripts (included)
pip install requests   # for citation_checker.py
python3 scripts/cross_ref_audit.py --help
python3 scripts/citation_checker.py --help

Usage

Invoke with /write-report in Claude Code, or describe your report needs and the skill will be suggested.

Quick start

1. Have your research repo with code + results ready
2. Tell Claude: “Write my FYP report based on this repo”
3. Claude will execute the 3-wave pipeline automatically

Citation verification

Verify all citations against 3 academic databases before submission. Catches AI-hallucinated references (6-55% of AI-generated citations are fabricated):

# Check all .bib files
python3 scripts/citation_checker.py path/to/report/

# JSON output for CI
python3 scripts/citation_checker.py references.bib --json

Cascading pipeline: CrossRef (140M+ DOIs) -> Semantic Scholar (200M+ papers) -> OpenAlex (240M+ works). Detects fully fabricated, chimeric (blended), and modified-real hallucinations.

Cross-reference audit

After parallel agents write chapters, catch duplicate labels:

python3 scripts/cross_ref_audit.py path/to/report/

What’s included

write-academic-report/
  SKILL.md                              # Main skill definition
  README.md                             # This file
  references/
    writing-guide.md                    # Academic writing philosophy
    citation-workflow.md                # Citation verification pipeline
    compilation-guide.md                # LaTeX compilation (tectonic, latexmk)
    parallel-pipeline.md                # Wave architecture documentation
  templates/