Paper Audit Skill (论文审核)
Unified academic paper auditing across formats and languages.
Critical Rules
- NEVER modify
\cite{}, \ref{}, \label{}, math environments in LaTeX
- NEVER modify
@cite, #cite(), #ref(), <label> in Typst
- NEVER fabricate bibliography entries — only verify existing
.bib/.yml files
- NEVER change domain terminology without user confirmation
- Check
FORBIDDEN_TERMS lists before suggesting any terminology changes
- For PDF input, clearly flag sections where extraction quality is uncertain
- Always distinguish between automated findings and LLM-judgment scores
Audit Modes
Mode: self-check (Pre-submission Self-Check)
Trigger keywords: audit, check, self-check, pre-submission, score, review my paper
What it does: Runs all automated checks and generates a structured report with:
- Per-dimension scores (Quality, Clarity, Significance, Originality) on 1-6 scale
- Issue list sorted by severity (Critical > Major > Minor)
- Improvement suggestions per section
- Pre-submission checklist results
CLI: python scripts/audit.py paper.tex --mode self-check
Mode: review (Peer Review Simulation)
Trigger keywords: simulate review, peer review, reviewer perspective, what would reviewers say
What it does: Everything in self-check PLUS:
- Paper summary from reviewer perspective
- Strengths analysis
- Weaknesses analysis with severity
- Questions a reviewer would ask
- Accept/reject recommendation with confidence
CLI: python scripts/audit.py paper.tex --mode review
Mode: gate (Quality Gate)
Trigger keywords: quality gate, pass/fail, can I submit, ready to submit, advisor check
What it does: Fast mandatory checks only:
- Format validation
- Bibliography integrity
- Figure/table references
- Pre-submission checklist
- Binary PASS/FAIL verdict with blocking issues
CLI: python scripts/audit.py paper.tex --mode gate
Supported Formats
| Format |
Parser |
Notes |
| LaTeX (.tex) |
LatexParser |
Full support — all checks available |
| Typst (.typ) |
TypstParser |
Full support — all checks available |
| PDF (.pdf) basic |
PdfParser (pymupdf) |
Text extraction with font-size heading detection |
| PDF (.pdf) enhanced |
PdfParser (pymupdf4llm) |
Structured Markdown with table/header preservation |
PDF Limitations: Math formulas may be lost; some checks (format, figures) skip for PDF. Recommend providing source files (.tex/.typ) for maximum accuracy.
Language Support
| Language |
Detection |
Extra Checks |
| English |
Auto (default) |
Standard suite |
| Chinese |
Auto (CJK ratio > 30%) |
+ consistency check, + GB/T 7714 compliance |
Force with --lang en or --lang zh.
Check Modules
| Module |
Script Source |
Dimensions Affected |
Applicable Formats |
| Format Check |
check_format.py |
Clarity |
.tex, .typ |
| Grammar Analysis |
analyze_grammar.py |
Clarity |
.tex, .typ, .pdf |
| Logic & Coherence |
analyze_logic.py |
Quality, Significance |
.tex, .typ, .pdf |
| Sentence Complexity |
analyze_sentences.py |
Clarity |
.tex, .typ, .pdf |
| De-AI Detection |
deai_check.py |
Clarity, Originality |
.tex, .typ, .pdf |
| Bibliography |
verify_bib.py |
Quality |
.tex, .typ |
| Figure/Table Refs |
check_figures.py |
Clarity |
.tex |
| Consistency (ZH) |
check_consistency.py |
Clarity |
.tex (Chinese only) |
| GB/T 7714 (ZH) |
verify_bib.py (GB mode) |
Quality |
.tex (Chinese only) |
| Pre-submission Checklist |
Built-in |
All |
All formats |
Scoring System
Based on REVIEWER_PERSPECTIVE.md criteria:
Four Dimensions
- Quality (30%): Technical soundness, well-supported claims
- Clarity (30%): Clear writing, reproducible, good organization
- Significance (20%): Community impact, advances understanding
- Originality (20%): New insights, not obvious extensions
Six-Point Scale (NeurIPS standard)
| Score |
Rating |
Meaning |
| 5.5-6.0 |
Strong Accept |
Groundbreaking, technically flawless |
| 4.5-5.4 |
Accept |
Technically solid, high impact |
| 3.5-4.4 |
Borderline Accept |
Solid but limited evaluation/novelty |
| 2.5-3.4 |
Borderline Reject |
Merits but weaknesses outweigh |
| 1.5-2.4 |
Reject |
Technical flaws, insufficient evaluation |
| 1.0-1.4 |
Strong Reject |
Fundamental errors or known results |
Output Protocol
All issues follow the unified format:
[MODULE] (Line N) [Severity: Critical|Major|Minor] [Priority: P0|P1|P2]: Issue description
Original: ...
Revised: ...
Rationale: ...
- Severity: Critical (must fix), Major (should fix), Minor (nice to fix)
- Priority: P0 (blocking), P1 (important), P2 (low priority)
Workflow
When a user requests a paper audit:
- Identify the file — locate the .tex, .typ, or .pdf file
- Determine mode — self-check (default), review, or gate based on user intent
- Run the orchestrator —
python scripts/audit.py <file> --mode <mode>
- Present the report — show the Markdown report to the user
- Discuss findings — help the user address Critical and Major issues first
- Re-audit if needed — run again after fixes to verify improvements
For review mode, supplement the automated report with LLM analysis of:
- Overall paper strengths (what works well)
- Key weaknesses (what reviewers would criticize)
- Questions a reviewer would ask
- Missing related work or baselines
1---2name: paper-audit3description: Unified paper audit skill supporting Chinese & English academic papers. Supports LaTeX (.tex), Typst (.typ), and PDF (.pdf) input formats. Three modes: self-check (pre-submission), review (peer review simulation), gate (quality gate pass/fail). Use when user mentions: audit, review, check paper, paper quality, pre-submission check, score paper, or any paper auditing task.4---56# Paper Audit Skill (论文审核)78Unified academic paper auditing across formats and languages.910## Critical Rules11121. NEVER modify `\cite{}`, `\ref{}`, `\label{}`, math environments in LaTeX132. NEVER modify `@cite`, `#cite()`, `#ref()`, `<label>` in Typst143. NEVER fabricate bibliography entries — only verify existing `.bib`/`.yml` files154. NEVER change domain terminology without user confirmation165. Check `FORBIDDEN_TERMS` lists before suggesting any terminology changes176. For PDF input, clearly flag sections where extraction quality is uncertain187. Always distinguish between automated findings and LLM-judgment scores1920## Audit Modes2122### Mode: `self-check` (Pre-submission Self-Check)2324**Trigger keywords**: audit, check, self-check, pre-submission, score, review my paper2526**What it does**: Runs all automated checks and generates a structured report with:27- Per-dimension scores (Quality, Clarity, Significance, Originality) on 1-6 scale28- Issue list sorted by severity (Critical > Major > Minor)29- Improvement suggestions per section30- Pre-submission checklist results3132**CLI**: `python scripts/audit.py paper.tex --mode self-check`3334### Mode: `review` (Peer Review Simulation)3536**Trigger keywords**: simulate review, peer review, reviewer perspective, what would reviewers say3738**What it does**: Everything in self-check PLUS:39- Paper summary from reviewer perspective40- Strengths analysis41- Weaknesses analysis with severity42- Questions a reviewer would ask43- Accept/reject recommendation with confidence4445**CLI**: `python scripts/audit.py paper.tex --mode review`4647### Mode: `gate` (Quality Gate)4849**Trigger keywords**: quality gate, pass/fail, can I submit, ready to submit, advisor check5051**What it does**: Fast mandatory checks only:52- Format validation53- Bibliography integrity54- Figure/table references55- Pre-submission checklist56- Binary PASS/FAIL verdict with blocking issues5758**CLI**: `python scripts/audit.py paper.tex --mode gate`5960## Supported Formats6162| Format | Parser | Notes |63|--------|--------|-------|64| LaTeX (.tex) | `LatexParser` | Full support — all checks available |65| Typst (.typ) | `TypstParser` | Full support — all checks available |66| PDF (.pdf) basic | `PdfParser` (pymupdf) | Text extraction with font-size heading detection |67| PDF (.pdf) enhanced | `PdfParser` (pymupdf4llm) | Structured Markdown with table/header preservation |6869**PDF Limitations**: Math formulas may be lost; some checks (format, figures) skip for PDF. Recommend providing source files (.tex/.typ) for maximum accuracy.7071## Language Support7273| Language | Detection | Extra Checks |74|----------|-----------|-------------|75| English | Auto (default) | Standard suite |76| Chinese | Auto (CJK ratio > 30%) | + consistency check, + GB/T 7714 compliance |7778Force with `--lang en` or `--lang zh`.7980## Check Modules8182| Module | Script Source | Dimensions Affected | Applicable Formats |83|--------|-------------|--------------------|--------------------|84| Format Check | `check_format.py` | Clarity | .tex, .typ |85| Grammar Analysis | `analyze_grammar.py` | Clarity | .tex, .typ, .pdf |86| Logic & Coherence | `analyze_logic.py` | Quality, Significance | .tex, .typ, .pdf |87| Sentence Complexity | `analyze_sentences.py` | Clarity | .tex, .typ, .pdf |88| De-AI Detection | `deai_check.py` | Clarity, Originality | .tex, .typ, .pdf |89| Bibliography | `verify_bib.py` | Quality | .tex, .typ |90| Figure/Table Refs | `check_figures.py` | Clarity | .tex |91| Consistency (ZH) | `check_consistency.py` | Clarity | .tex (Chinese only) |92| GB/T 7714 (ZH) | `verify_bib.py` (GB mode) | Quality | .tex (Chinese only) |93| Pre-submission Checklist | Built-in | All | All formats |9495## Scoring System9697Based on REVIEWER_PERSPECTIVE.md criteria:9899### Four Dimensions100- **Quality** (30%): Technical soundness, well-supported claims101- **Clarity** (30%): Clear writing, reproducible, good organization102- **Significance** (20%): Community impact, advances understanding103- **Originality** (20%): New insights, not obvious extensions104105### Six-Point Scale (NeurIPS standard)106| Score | Rating | Meaning |107|-------|--------|---------|108| 5.5-6.0 | Strong Accept | Groundbreaking, technically flawless |109| 4.5-5.4 | Accept | Technically solid, high impact |110| 3.5-4.4 | Borderline Accept | Solid but limited evaluation/novelty |111| 2.5-3.4 | Borderline Reject | Merits but weaknesses outweigh |112| 1.5-2.4 | Reject | Technical flaws, insufficient evaluation |113| 1.0-1.4 | Strong Reject | Fundamental errors or known results |114115## Output Protocol116117All issues follow the unified format:118119```120[MODULE] (Line N) [Severity: Critical|Major|Minor] [Priority: P0|P1|P2]: Issue description121 Original: ...122 Revised: ...123 Rationale: ...124```125126- **Severity**: Critical (must fix), Major (should fix), Minor (nice to fix)127- **Priority**: P0 (blocking), P1 (important), P2 (low priority)128129## Workflow130131When a user requests a paper audit:1321331. **Identify the file** — locate the .tex, .typ, or .pdf file1342. **Determine mode** — self-check (default), review, or gate based on user intent1353. **Run the orchestrator** — `python scripts/audit.py <file> --mode <mode>`1364. **Present the report** — show the Markdown report to the user1375. **Discuss findings** — help the user address Critical and Major issues first1386. **Re-audit if needed** — run again after fixes to verify improvements139140For `review` mode, supplement the automated report with LLM analysis of:141- Overall paper strengths (what works well)142- Key weaknesses (what reviewers would criticize)143- Questions a reviewer would ask144- Missing related work or baselines
tools-only