name: pdf-rename description: Rename academic PDF papers to a standardized format "[Year] [Venue] Title.pdf" using a three-stage pipeline (Extract → Verify → Rename). Use when the user asks to organize, batch-rename, or metadata-enrich PDF files in a folder. Activates on keywords like "rename PDFs", "organize papers", "batch rename PDFs", "rename papers by metadata", "pdf重命名", "文献整理".

PDF Rename — Academic Paper Organizer

Rename academic PDFs to: [Year] [Venue] Title.pdf

Three-stage pipeline:

Extract → Verify → Rename

Anti-error principle: Never re-parse PDF content during Rename. The Manifest is the single source of truth.

Quick Start

# Stage 1: Extract raw text → manifest.json
python scripts/extract.py "<folder_path>"

# Stage 2: LLM parses raw_text → inject verified data → manifest_verified.json
#   (Agent reads manifest.json raw_text field and writes to scripts/VERIFIED_DATA_*.py)
python scripts/apply_verified.py "<folder_path>"

# Stage 3: Preview / Execute
python scripts/execute.py "<folder_path>" --preview
python scripts/execute.py "<folder_path>" --execute

Stage 1: Extract

What it does: - Reads first 3 pages of each PDF - Stores raw text in manifest.json → raw_text field - Extracts year_hint from filename prefix - Detects potential duplicates by filename similarity

Manifest schema — see references/manifest_spec.md

Stage 2: Verify

The agent reads manifest.json, parses each raw_text field, and writes verified metadata.

Steps: 1. Read manifest.json 2. For each entry, parse the raw_text to extract: title, authors, venue, year, abstract 3. Create or update scripts/VERIFIED_DATA_*.py with verified entries

VERIFIED_DATA format:

VERIFIED_DATA = {
    "OriginalFilename.pdf": {
        "title": "Correct Paper Title",
        "year": "2024",
        "venue": "NeurIPS",
        "confirmed": True   # must be True to be renamed
    },
}

Rules: - Key must exactly match the original filename - confirmed: True → status becomes ready → will be renamed - confirmed: False or omitted → skipped - Multiple VERIFIED_DATA_*.py files are auto-merged - Prefer venue/conference year over filename year (e.g., arXiv 2022 → NeurIPS 2024)

⚠️ Key gotchas during parsing: - Multi-line titles: pypdf concatenates lines without spaces (e.g., "Direct score maximization outperformsplanning loss") — use context to infer correct split - Journal refs at top of page: "Math. Program. (2012) 133..." is NOT the title — read further - arXiv papers: use conference year if published, otherwise use arXiv year - Non-paper files (book chapters, lecture notes, loss curves): set confirmed: False and skip

Then run:

python scripts/apply_verified.py "<folder_path>"

Stage 3: Rename

• Only status == 'ready' files are renamed
• Duplicate titles → (1), (2) suffixes
• Backup created at <folder>/_backup_YYYYMMDD_HHMMSS/

python scripts/execute.py "<folder_path>" --preview   # dry run
python scripts/execute.py "<folder_path>" --execute   # rename

Scripts

Known Failure Modes

7w4.net小葱技能站，你的AI助手技能库。

References

• references/manifest_spec.md — Full manifest JSON schema
• references/venue_abbrev.md — Standard venue abbreviation map

这个工具质量不错，能够有效整理学术论文 PDF，按年份、期刊/会议、标题重新命名。防错机制很用心，预览重命名、备份恢复、确认后才执行，三重保护让人安心。文档写得很详细，新手也能快速上手。不过工具包里有残留的测试数据比较碍眼，而且跨平台使用（比如 Linux/Mac）可能会有路径问题。整体适合有大量论文需要整理的用户，但中文文件名支持一般。.