From buaa-thesis-checker
学术论文内容审核工具。用于审核论文的主要贡献、方法创新性、实验评估及baseline对比。触发:用户提到"审核论文"、"论文评审"、"paper review"、"论文创新性"、"论文贡献"、"论文评估"、"review paper"、"audit paper"、"论文对比baseline"。核心:追溯关系图(P→I→E)、CCF顶会顶刊判断、贡献点数量约束(≤3个)。输出:Markdown + HTML格式中文英文审核报告。
How this skill is triggered — by the user, by Claude, or both
Slash command
/buaa-thesis-checker:buaa-thesis-content-checkingThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Activate when user mentions: "审核论文"、"论文评审"、"paper review"、"论文创新性"、"论文贡献"、"论文评估"、"review paper"、"audit paper"、"论文对比baseline"
README.mdevals/evals.jsonreferences/paper_audit_template.htmlscripts/check_deps.pyscripts/checks/__init__.pyscripts/checks/baseline_comparison.pyscripts/checks/contributions.pyscripts/checks/experiments.pyscripts/checks/innovation.pyscripts/extractors/__init__.pyscripts/extractors/content.pyscripts/extractors/llm_extractor.pyscripts/paper_audit_script.pyscripts/reports/__init__.pyscripts/reports/generator.pyscripts/reports/template_filler.pyActivate when user mentions: "审核论文"、"论文评审"、"paper review"、"论文创新性"、"论文贡献"、"论文评估"、"review paper"、"audit paper"、"论文对比baseline"
# Step 1: Check and install dependencies automatically
python3 <skill_path>/scripts/check_deps.py --yes
# Step 2: Run PDF extraction and analysis
python3 <skill_path>/scripts/paper_audit_script.py --step1 <pdf_path> [output_dir]
IMPORTANT: Before running the script, always check that required packages are installed:
python3 -c "import pdfplumber; import fitz; from pypdf import PdfReader; print('All dependencies OK')"
If this fails, install dependencies with:
python3 -m pip install pymupdf pdfplumber pypdf --quiet
Or use the helper script:
python3 <skill_path>/scripts/check_deps.py --yes
Use --check-only when you only want dependency status. In non-interactive shells, do not call check_deps.py without --yes or --check-only.
主要问题 (P) → 方法创新 (I) → 评估验证 (E)
| Task | Description | Dependency |
|---|---|---|
| #1 | PDF text extraction | - |
| #2 | Reference investigation (venue, year, CCF level) | #1 |
| #3 | LLM comprehensive analysis → JSON | #1, #2 |
| #4 | Report generation (Markdown + HTML zh/en) | #3 |
Task 1 — PDF Extraction:
# First ensure dependencies are installed
python3 <skill_path>/scripts/check_deps.py --yes
# Then run extraction
python3 <skill_path>/scripts/paper_audit_script.py --step1 <pdf_path> [output_dir]
Task 2 — Reference Investigation: For each method/issue found, search for its original paper and verify:
Task 3 — LLM Analysis (JSON output): Analyze paper and produce structured JSON with:
Task 4 — Report Generation (Agent): Generate 3 files from JSON:
paper_audit_report_YYYYMMDD_HHMMSS.md — Chinese Markdownpaper_audit_report_YYYYMMDD_HHMMSS_zh.html — Chinese HTMLpaper_audit_report_YYYYMMDD_HHMMSS_en.html — English HTMLIf the full automatic content report fails because anthropic is unavailable, ANTHROPIC_API_KEY is not set, or the LLM/report script raises an exception:
--step1 output, especially paper_text_extracted.txt.When the PDF is scanned/image-based and text extraction is poor, stop and request OCR before judging academic content.
| Issue | Solution |
|---|---|
ModuleNotFoundError: No module named 'pdfplumber' | Run python3 -m pip install pdfplumber pymupdf pypdf |
EOFError in check_deps.py | Run python3 scripts/check_deps.py --yes or --check-only |
ModuleNotFoundError: No module named 'anthropic' | Continue with extracted text if no API-based LLM report is required; install anthropic only when automatic LLM extraction is needed |
ModuleNotFoundError: name 'Dict' is not defined | The script already includes from typing import Dict, List, Optional, Any |
NameError: name 're' is not defined | Update to the current script version; fallback analysis depends on top-level import re |
| PDF is scanned/image-based | Use OCR tool first, then feed text to the script |
CCF-A Conferences: NeurIPS, ICML, ICLR, ACL, EMNLP, AAAI, IJCAI, ASE, ICSE, FSE, ISSTA, OOPSLA, PLDI, POPL, SIGIR, KDD, WWW, CHI CCF-A Journals: IEEE TPAMI, IJCV, JMLR, ACM Computing Surveys, IEEE TNNLS, IEEE TKDE, IEEE TSE
| Type | Max | If Exceeded |
|---|---|---|
| Evaluation metrics | 3 | Merge related metrics |
| Problems | 3 | Merge root-cause related |
| Innovations | 3 | Merge method-similar |
| Contributions | 3 | Refine to core contributions |
python3 -m pip install pymupdf pdfplumber pypdf
# Optional only for API-backed automatic LLM extraction:
python3 -m pip install anthropic
references/CCF-List.mdCreates, edits, and optimizes skills for Claude Code, including drafting, evaluating with test prompts, iterating on performance, and improving skill descriptions for better triggering accuracy.
npx claudepluginhub yylonly/buaa-thesis-checker --plugin buaa-thesis-checker