Post-writing polish and validation for academic papers — format checks, grammar analysis, de-AI editing, bibliography verification, and experiment narrative generation. Focused on enhancing existing text quality, not generating from scratch.
Recommended platforms: Claude Code · Codex · Antigravity
| Skill | Best For | Input Formats |
|---|---|---|
latex-paper-en |
English papers — IEEE / ACM / NeurIPS / ICML / Springer | .tex |
latex-thesis-zh |
Chinese degree theses — GB/T 7714 / thuthesis / pkuthss | .tex |
typst-paper |
Fast-compile bilingual papers | .typ |
paper-audit |
Deep-review-first paper audit and submission gate | .tex .typ .pdf |
industrial-ai-research |
Industrial AI literature synthesis & gap analysis | — |
Install via skills, the community skill manager for Claude Code:
# Install individual skills
npx skills add github.com/bahayonghang/academic-writing-skills/latex-paper-en
npx skills add github.com/bahayonghang/academic-writing-skills/latex-thesis-zh
npx skills add github.com/bahayonghang/academic-writing-skills/typst-paper
npx skills add github.com/bahayonghang/academic-writing-skills/paper-audit
npx skills add github.com/bahayonghang/academic-writing-skills/industrial-ai-research
# Or install everything at once
npx skills add github.com/bahayonghang/academic-writing-skillsgit clone https://github.com/bahayonghang/academic-writing-skills.git
cd academic-writing-skills/academic-writing-skillsLinux / macOS
mkdir -p ~/.claude/skills
cp -r latex-paper-en latex-thesis-zh typst-paper paper-audit industrial-ai-research ~/.claude/skills/Windows (PowerShell)
New-Item -ItemType Directory -Path "$env:USERPROFILE/.claude/skills" -Force
foreach ($skill in @("latex-paper-en","latex-thesis-zh","typst-paper","paper-audit","industrial-ai-research")) {
Copy-Item -Recurse $skill "$env:USERPROFILE/.claude/skills/"
}English academic paper editing for IEEE, ACM, Springer, NeurIPS, and ICML venues.
| Category | Capability |
|---|---|
| Format & Compile | ChkTeX linting; pdfLaTeX / XeLaTeX / LuaLaTeX via latexmk |
| Grammar | Chinglish detection, weak-verb replacement, subject-verb agreement |
| Sentences | Complex sentence decomposition (auto-triggers at > 50 words) |
| Expression | Academic tone improvement, expression restructuring |
| Logic | Paragraph coherence (AXES model), introduction funnel checks, methodological depth, abstract/conclusion alignment |
| Title | IEEE/ACM/Springer best-practice generation; removes filler words; scores 0–100 |
| Captions | Title/Sentence-case, AI-flavor-free figure and table captions |
| Pseudocode | IEEE-safe review for algorithm2e, algorithmicx, algpseudocodex; checks float usage, caption/label/reference hygiene, long comments, and advisory line-number defaults |
| Experiments | Cohesive result paragraphs with SOTA comparison, ablation analysis, discussion layering, and conclusion completeness |
| De-AI | Humanize AI-written passages while preserving all LaTeX syntax; flags low-information boilerplate |
| Anti-Citation-Stacking | Max 2 clustered citations per sentence; flags stacking in Introduction/Related Work |
| References | Undefined \ref{}, unreferenced \label{}, missing captions; BibTeX format validation |
| Online Verify | CrossRef + Semantic Scholar bibliography verification (no API key required) |
| Translation | Chinese → English academic translation with domain-term awareness |
Chinese degree thesis editing conforming to GB/T 7714-2015 and major university templates.
| Category | Capability |
|---|---|
| Structure | Multi-file thesis mapping; chapter/section completeness check |
| Bibliography | GB/T 7714-2015 compliance; BibTeX format validation |
| Templates | thuthesis / pkuthss / ustcthesis / fduthesis auto-detection |
| Chinese Style | Oral-expression detection, terminology consistency |
| Logic | Paragraph coherence (AXES model), introduction funnel, chapter mainline checks, cross-section closure |
| Title | GB/T 7713.1-2006 compliant; Chinese & English bilingual candidates |
| Captions | Bilingual English/Chinese captions following top-conference standards |
| Experiments | Chinese core-journal narrative paragraphs with baseline/ablation coverage, discussion layering, and conclusion completeness |
| De-AI | Reduce AI writing traces; preserves all LaTeX commands; flags low-information rhetoric |
| Anti-Citation-Stacking | Max 2 clustered citations per sentence; flags stacking in Introduction/Related Work chapters |
| Compile | XeLaTeX / LuaLaTeX with full CJK font support |
| References | Same integrity checks as latex-paper-en; online verification supported |
Supported university templates
| University | Template | Notes |
|---|---|---|
| Tsinghua | thuthesis | Figure numbering: 图 3-1 |
| Peking | pkuthss | Nomenclature chapter required |
| USTC | ustcthesis | — |
| Fudan | fduthesis | — |
| Generic | ctexbook | GB/T 7713.1-2006 baseline |
Bilingual Typst paper editing with millisecond-level compilation.
| Category | Capability |
|---|---|
| Compile | Typst CLI wrapper with error summarization |
| Format | Page settings, text formatting, citation syntax |
| Grammar | Same checks as latex-paper-en, adapted for Typst syntax |
| Logic | AXES paragraph coherence, introduction funnel, abstract/conclusion alignment, cross-section closure |
| Title | Bilingual (English/Chinese) title generation and optimization |
| Captions | Bilingual captions following IEEE/ACM standards |
| Pseudocode | IEEE-like review for algorithmic, algorithm-figure, and lovelace, including wrapper, caption, style hook, and comment-length checks |
| Experiments | Cohesive result paragraphs for journal/conference papers, including discussion layering checks |
| De-AI | Humanize AI-written passages in English or Chinese; preserves @cite, <label>, $...$ |
| Anti-Citation-Stacking | Max 2 clustered citations per sentence; flags stacking in Introduction/Related Work |
| Venues | IEEE, ACM, Springer, NeurIPS template guidance |
| References | Undefined @ref, unreferenced labels; online verification supported |
Deep-review-first paper audit with layered checks, structured issue bundles, and submission gating.
| Category | Capability |
|---|---|
| Input | .tex, .typ, .pdf files |
| Modes | quick-audit (fast screen) · deep-review (reviewer-style critique) · gate (submission gate) · re-audit (revision verification) |
| Visual Layout | Margin overflow, text/image overlaps, font inconsistency, low-res images, blank pages |
| Reference Integrity | Undefined refs, unreferenced labels, missing captions, numbering gaps |
| Caption Audit | Title/Sentence case enforcement; AI-flavor removal |
| Pseudocode Audit | IEEE gate checks for floating algorithm environments, caption/label/reference hygiene, plus advisory checks for line numbers and long comments |
| Experiment Narrative | Checks paragraph cohesion, baseline comparisons, discussion depth/layering, and conclusion completeness |
| Deep Review Outputs | final_issues.json, overall_assessment.txt, review_report.md, revision_roadmap.md |
| ScholarEval | 8-dimension quality scoring (1–10) with publication readiness label |
| NeurIPS Scoring | Quality / Clarity / Significance / Originality on 1–6 scale |
| Online Verify | CrossRef + Semantic Scholar (add --online); no API key required |
| De-AI | Reduce AI writing traces across the whole document |
| Citation Stacking | Detects 3+ clustered citations without individual discussion in Introduction/Related Work |
| Review Scope Note | Phase 0 is script-backed; deep-review adds quote-anchored reviewer lanes for claims-vs-evidence, notation/numeric consistency, evaluation fairness, self-consistency, and prior-art grounding |
Audit workflow layers
| Layer | Check |
|---|---|
| L0 | Quick audit / gate script pass |
| L1 | Deep-review workspace prep (sections, summary, claim map) |
| L2 | Section review lanes |
| L3 | Cross-cutting review lanes |
| L4 | Consolidation + quote verification |
| L5 | Final report + roadmap + optional score summary |
Quick usage
| Mode | Use when | Main output |
|---|---|---|
quick-audit |
You want a fast readiness screen | Script-backed report + checklist + score summary |
deep-review |
You want reviewer-style critique | Structured issue bundle + roadmap |
gate |
You only care about blockers | PASS/FAIL + blocking issues |
re-audit |
You want to verify revisions | Issue-status comparison |
uv run python academic-writing-skills/paper-audit/scripts/audit.py paper.tex --mode quick-audit
uv run python academic-writing-skills/paper-audit/scripts/audit.py paper.tex --mode deep-review --scholar-eval
uv run python academic-writing-skills/paper-audit/scripts/audit.py paper.tex --mode gate --venue ieee
uv run python academic-writing-skills/paper-audit/scripts/audit.py paper.tex --mode re-audit --previous-report report_v1.mdCompatibility aliases:
self-check->quick-auditreview->deep-review
Docs:
Structured literature synthesis focused on Industrial AI domains.
| Category | Capability |
|---|---|
| Domains | Predictive maintenance, intelligent scheduling, anomaly detection, smart manufacturing, CPS, robotics |
| Intake | Asks for report language, deliverable mode, time window, and emphasis before synthesizing |
| Retrieval | Prioritizes recent arXiv + top IEEE/automation venues (T-ASE, CASE, T-II) |
| Outputs | research-brief · literature-map · venue-ranked survey · research-gap memo · survey-draft |
| Survey Draft | Taxonomy-driven outline → per-section evidence packs → section-by-section writing → merge with quality gate; optional LaTeX handoff |
| Report Structure | Search scope → source buckets → shortlisted papers → synthesis → next-step recommendations |
Skills auto-trigger from natural language. Just open Claude Code and describe your task.
compile my paper with xelatex-biber
build the LaTeX document
| Recipe | Steps | Use Case |
|---|---|---|
xelatex |
XeLaTeX only | Quick Chinese compile |
pdflatex |
PDFLaTeX only | Quick English compile |
latexmk |
LaTeXmk auto | Auto dependency handling |
xelatex-biber |
xelatex → biber → xelatex × 2 | Chinese + Biber (Recommended) |
xelatex-bibtex |
xelatex → bibtex → xelatex × 2 | Chinese + BibTeX |
pdflatex-biber |
pdflatex → biber → pdflatex × 2 | English + Biber |
pdflatex-bibtex |
pdflatex → bibtex → pdflatex × 2 | English + BibTeX |
deai check my introduction
humanize this paragraph
reduce AI writing traces in Section 3
Removes empty phrases, over-confident expressions, and mechanical parallel structures. Preserves all LaTeX/Typst commands.
check grammar in abstract
improve academic tone in the related work section
detect Chinglish in Section 2
check logical coherence in my introduction
analyze methodological depth
verify paragraph structure with AXES model
optimize my paper title
generate 5 title candidates for this paper
Follows IEEE/ACM/Springer/NeurIPS best practices. Removes ineffective words ("Novel", "A Study of", "Research on"). Scores candidates 0–100. Ensures key terms appear in the first 65 characters (English) / 20 characters (Chinese).
analyze my experiment data and write results section
generate ablation study analysis paragraph
write a SOTA comparison paragraph from this table
Output: cohesive narrative paragraphs (LaTeX/Typst), not bullet lists.
generate IEEE-standard figure captions
optimize this table caption
generate bilingual caption for Figure 3
check whether this IEEE pseudocode still uses algorithm2e floats
review this algorithm-figure block for caption and line-number issues
make this pseudocode IEEE-safe without inventing a fake Algorithm 1 rule
check figure references in my paper
find undefined labels
verify my bibliography
run a full audit on my paper
check paper quality before submission
audit my PDF for layout issues
run paper-audit --online --scholar-eval
translate this section to English
中译英这个段落
Auto-detects domain terminology (Deep Learning, Time Series, Industrial Control).
All suggestions use diff-comment style with mandatory severity and priority fields:
% <MODULE> (Line <N>) [Severity: Critical|Major|Minor] [Priority: P0|P1|P2]: <Issue summary>
% Before: <original text>
% After: <suggested text>
% Rationale: <brief explanation>
% ⚠️ [PENDING VERIFICATION]: <if evidence/metric is required>| Severity | Meaning |
|---|---|
| Critical | Blocks submission (compilation failure, undefined reference, missing required section) |
| Major | Significantly affects quality (grammar error, logic gap, non-compliant format) |
| Minor | Polish-level improvement (word choice, style consistency) |
- Python 3.10+
- TeX Live or MiKTeX (with
latexmk,chktex) - Chinese documents: XeLaTeX + CJK fonts (SimSun, SimHei, KaiTi)
- Python 3.10+
- Typst CLI (
cargo install typst-clior system package manager) - Chinese documents: Source Han Serif / Noto Serif CJK SC
- Python 3.10+
pdfplumberfor PDF visual analysis (uv syncorpip install pdfplumber)
academic-writing-skills/
├── latex-paper-en/
│ ├── SKILL.md # Skill entry point & trigger keywords
│ ├── agents/ # Agent metadata
│ ├── evals/ # Evaluation cases
│ ├── examples/ # Example prompts
│ ├── references/ # Style guides, venue rules, forbidden terms
│ └── scripts/
│ ├── parsers.py # LatexParser / TypstParser base
│ ├── compile.py # Unified compiler (pdflatex/xelatex/latexmk)
│ ├── check_format.py # ChkTeX wrapper
│ ├── verify_bib.py # BibTeX format validation
│ ├── online_bib_verify.py # CrossRef / Semantic Scholar lookup
│ ├── check_references.py # \ref / \label / caption integrity
│ ├── check_figures.py # Figure usage analysis
│ ├── check_pseudocode.py # IEEE-aware pseudocode checks
│ ├── analyze_grammar.py # Chinglish, weak verbs, agreement
│ ├── analyze_sentences.py # Long sentence decomposition
│ ├── analyze_logic.py # AXES coherence, transition signals
│ ├── improve_expression.py # Academic tone restructuring
│ ├── optimize_title.py # Title generation & scoring
│ ├── analyze_experiment.py # Experiment narrative generation
│ ├── deai_check.py # Single-passage de-AI
│ ├── deai_batch.py # Batch de-AI over full document
│ ├── translate_academic.py # CN→EN domain-aware translation
│ └── extract_prose.py # Plain-text extraction (skip math/env)
│
├── latex-thesis-zh/
│ ├── SKILL.md
│ ├── agents/ · evals/ · examples/ · references/
│ └── scripts/ # Same toolset as latex-paper-en plus:
│ ├── map_structure.py # Multi-file thesis structure mapper
│ ├── detect_template.py # Template auto-detection
│ └── check_consistency.py # Terminology & notation consistency
│
├── typst-paper/
│ ├── SKILL.md
│ ├── agents/ · evals/ · examples/
│ ├── references/ # STYLE_GUIDE.md, TYPST_SYNTAX.md, DEAI_GUIDE.md
│ └── scripts/ # Same toolset, Typst-syntax aware
│ └── check_pseudocode.py # IEEE-like Typst pseudocode checks
│
├── paper-audit/
│ ├── SKILL.md
│ ├── agents/ · examples/ · templates/
│ ├── references/
│ │ └── SCHOLAR_EVAL_GUIDE.md
│ └── scripts/
│ ├── audit.py # Main orchestrator
│ ├── parsers.py # Shared parser base
│ ├── pdf_parser.py # PDF text & metadata extraction
│ ├── visual_check.py # PDF layout & rendering analysis
│ ├── check_pseudocode.py # Routed from sibling skills for IEEE pseudocode checks
│ ├── check_references.py # Reference integrity
│ ├── detect_language.py # Language detection
│ ├── scholar_eval.py # 8-dimension ScholarEval scoring
│ └── report_generator.py # Structured audit report output
│
└── industrial-ai-research/
├── SKILL.md
├── agents/ · examples/
└── references/ # Source policy, venue priority list
| Problem | Fix |
|---|---|
| Missing LaTeX tools | Install TeX Live / MiKTeX; ensure latexmk and chktex are in PATH |
| Missing Typst CLI | cargo install typst-cli or install via package manager |
| Compilation error | Summarize the first error block and request the relevant .log snippet |
| Missing script | Verify working directory points to the skill's root folder |
| PDF analysis fails | Install pdfplumber (uv sync --extra dev) |
Issues and pull requests are welcome. Please keep changes scoped to the relevant skill and run just ci before submitting.
Academic Use Only — Not for commercial use.
Full documentation is available at the docs directory.
Run locally:
cd docs
npm install
npm run docs:dev
# Open http://localhost:5173