yuan1z0825/nature-skillsCheck passed
SKILL DETAIL
nature-citation
yuan1z0825/nature-skills/nature-citation
The nature-citation skill is designed to add strict Nature/CNS citations to academic writing. It splits long passages into citable segments, searches only accepted flagship and subjournal titles from Nature Portfolio, the AAAS Science family, and Cell Press, and filters by publication time range. By default, it exports one reference-manager-ready output (e.g., RIS, EndNote, Zotero RDF) and ensures complete structured author metadata. It is suitable for users who need automatic citation insertion, segmented citation, Nature-series or CNS and subjournal support, reference supplementation, or reference export.
Installs · 146View source
Installation
npx skills add https://github.com/yuan1z0825/nature-skills --skill nature-citation
Skill files
SKILL.md
Last synced · Aug 27, 2026
agents/openai.yaml›
interface:
display_name: "Nature Citation"
short_description: "Find and verify Nature and CNS supporting citations"
default_prompt: "Use $nature-citation to add verified Nature and CNS citations to this manuscript passage."
evals/evals.json›
{
"skill_name": "nature-citation",
"evals": [
{
"id": 1,
"prompt": "把这段文字自动分段并给出Nature/CNS及其子刊引用，导出Zotero RDF格式和HTML可视化：Tumor-associated macrophages promote immune evasion by suppressing cytotoxic T cell activity. Single-cell RNA sequencing reveals cellular heterogeneity in pancreatic cancer.",
"expected_output": "Segments the text, maps each segment to citation candidates, exports references.rdf, and provides an HTML visualization that can download selected references as ENW, RIS, or Zotero RDF.",
"files": []
},
{
"id": 2,
"prompt": "只看Nature系列，把下面这一段按长度分段，给我文本和引用的对应关系，方便插入论文：single-cell RNA sequencing reveals cellular heterogeneity in pancreatic cancer. Spatial transcriptomics further preserves tissue context for interpreting tumor microenvironments.",
"expected_output": "Restricts scope to Nature Portfolio-style journals and produces a segment-reference correspondence table.",
"files": []
},
{
"id": 3,
"prompt": "Find flagship Nature/Science/Cell references for: CRISPR screens can identify genetic dependencies in cancer cells. Export RIS for EndNote.",
"expected_output": "Restricts to Nature, Science, and Cell only, treats the claim as a segment, and exports RIS without fabricating missing metadata.",
"files": []
},
{
"id": 4,
"prompt": "给我一个Nature系列引用导出，用户要自己选择下载 ENW、RIS 还是 Zotero RDF，并且先按年份筛选再勾选参考文献。",
"expected_output": "Produces the citation browser HTML with year filters, selectable references, and downloadable ENW/RIS/Zotero RDF exports from the same page.",
"files": []
}
]
}
manifest.yaml›
name: nature-citation
version: 2.1.0
description: >
Declarative manifest for the static/dynamic split. SKILL.md uses this to
decide which fragments to load for a citation request.
# Design note: nature-citation is a linear, parameterised workflow (segment ->
# parse -> search -> evaluate -> export -> report). Its variation — journal
# scope, user language, input length — is handled at runtime by inline rules and
# by the nature_citation.py flags (--scope, --batch-size, ...), not by loading
# different large content fragments. There is therefore no content axis; the
# split is core (always loaded) plus on-demand references. Heavy, only-sometimes-
# needed material (the script flag reference and long-article batching strategy)
# lives in references so a normal short run stays cheap. nature-citation does not
# use the prose-oriented nature-shared layer.
always_load:
- static/core/principles.md
- static/core/chinese-mode.md
- static/core/workflow.md
references:
on_demand:
- condition: running nature_citation.py — full flag list, long-article batch strategy, quick-guide tables
path: references/script-usage.md
- condition: translating a manuscript claim into search queries and support grades
path: references/search-strategy.md
- condition: the default Nature/CNS journal-family boundary and official source notes
path: references/journal-scope.md
- condition: RIS, EndNote, or Zotero RDF export guidance
path: references/ris-endnote.md
- condition: segment text, search Crossref, export ENW/RIS/RDF, and generate the HTML browser
path: scripts/nature_citation.py
README_EN.md›
# `nature-citation` Skill
[中文说明](README.md)
`nature-citation` splits manuscript passages or scientific claims into citable units and finds supporting references from Nature Portfolio, the Science family, and Cell Press.
## What To Use It For
- Add citations to key claims in an introduction, discussion, or reviewer response.
- Split long passages into stable claim units such as `S001` and `S002`.
- Restrict evidence to Nature, Science, Cell, and their subjournals, or keep only flagship journals.
- Explain each candidate paper's support position, evidence strength, and insertion point.
- Export inspectable RIS by default and block records with missing, surname-only, or otherwise incomplete author metadata before export.
- Retrieve structured authors by DOI or PMID while retaining order, given names or initials, suffixes, and collective authors.
## Typical Requests
- "Split this introduction paragraph and add Nature-series citations."
- "Use only CNS and subjournal papers from the last five years to support these claims."
- "I have confirmed these DOIs; export them in a Zotero-importable format."
## What You Need To Provide
- Passage to cite, claim list, DOI list, or PMID list.
- Journal scope, year range, whether reviews are allowed, and whether only flagship journals should be kept.
- Target citation style and export format such as `RIS`, `ENW`, or Zotero `RDF`.
## Outputs
- Claim-segmentation table and candidate-reference table.
- Suggested insertion position, DOI, journal, year, and support note for each claim.
- Optional JSON, TSV, Markdown, or HTML review materials.
- Reference-management export file. RIS is the default; choose `Reference Manager (RIS)` when importing it into EndNote.
## Boundaries
- Candidate papers are support options, not a guarantee that the final citation is appropriate.
- Blogs, press releases, and search snippets are not used as sole evidence.
- When a paper supports a nearby but not identical claim, the evidence mismatch is stated explicitly.
## Related Skills
- `nature-academic-search`: broader literature search and citation-metric audits.
- `nature-ref-verifier`: verify selected reference metadata.
- `nature-writing`: integrate citation choices back into manuscript argument.
README.md›
# `nature-citation` 技能
[English](README_EN.md)
`nature-citation` 用于把论文段落、手稿片段或单条科学判断拆成可引用的 claim，并为每个 claim 寻找 Nature Portfolio、Science family 和 Cell Press 范围内的支撑文献。
## 适合用它做什么
- 给 introduction、discussion 或 reviewer response 中的关键判断补引用。
- 将长段落拆成稳定编号的 claim 单元，例如 `S001`、`S002`。
- 限定只查 Nature、Science、Cell 及其子刊，或只保留旗舰刊。
- 为每个候选文献说明支撑位置、证据强度和插入建议。
- 默认导出可检查的 RIS 文件，并在导出前拦截缺失名字、仅有姓氏或作者列表不完整的记录。
- 通过 DOI 或 PMID 获取结构化作者信息，保留作者顺序、名字/首字母、后缀和团体作者。
## 典型请求
- “把这段 introduction 分段补 Nature 系列引用。”
- “只用 CNS 及子刊，为这些 claim 找近五年的支撑文献。”
- “我已经确认这些 DOI，帮我导出 Zotero 可导入文件。”
## 你需要提供
- 待引用的段落、claim 列表、DOI 清单或 PMID 清单。
- 期刊范围、时间范围、是否允许综述、是否只要旗舰刊。
- 目标引用格式和导出格式，例如 `RIS`、`ENW` 或 Zotero `RDF`。
## 产出
- claim 分段表和候选文献表。
- 每个 claim 的建议插入位置、候选 DOI、期刊、年份和支撑说明。
- 可选的 JSON、TSV、Markdown、HTML 审查材料。
- 可导入文献管理器的引用文件。默认使用 RIS，EndNote 导入时选择 `Reference Manager (RIS)`。
## 边界
- 只把论文作为候选支撑，不会替作者保证其必然适合最终引用。
- 不会使用博客、新闻稿或搜索摘要作为唯一依据。
- 当文献只能支撑相邻但不完全相同的 claim 时，会明确标注证据偏差。
## 相关技能
- `nature-academic-search`：更宽范围的文献搜索和引用指标审计。
- `nature-ref-verifier`：校验已选参考文献的元数据。
- `nature-writing`：把引用选择整合回手稿论证。
references/journal-scope.md›
# Journal Scope
The skill's default journal-family boundary is intentionally practical rather than exhaustive. Use it
to find likely Nature/CNS-family candidates, then verify exact journal status on official pages if the
author needs a strict portfolio definition.
## Default families
### Nature Portfolio
Include:
- `Nature`
- journals beginning with `Nature `, such as `Nature Medicine`, `Nature Biotechnology`,
`Nature Methods`, `Nature Materials`, `Nature Genetics`, `Nature Communications`
- `Communications` journals, such as `Communications Biology`, `Communications Chemistry`,
`Communications Materials`, `Communications Earth & Environment`, `Communications Medicine`
- `npj` journals
- `Scientific Reports`
Be careful with unrelated titles that include the common word "nature".
### Science family
Include by default:
- `Science`
- `Science Advances`
- `Science Translational Medicine`
- `Science Signaling`
- `Science Immunology`
- `Science Robotics`
The AAAS Science Partner Journal program is not included by default unless the user asks for partner
journals or broader AAAS coverage.
### Cell Press
Include the flagship `Cell`, major primary-research Cell Press journals, Cell Reports titles, and
Trends review journals. The local script recognizes common Cell Press titles and any title beginning
with `Trends in `.
Because Cell Press launches and reorganizes titles over time, verify official pages for exhaustive
coverage or a current journal list.
## Flagship-only scope
Use only:
- `Nature`
- `Science`
- `Cell`
This is appropriate when the user says "只看正刊", "主刊", "flagship only", or explicitly excludes
subjournals.
## Official source notes
- Crossref REST API can retrieve scholarly metadata, search works, and filter exact fields such as
`container-title` and `issn`.
- NCBI E-utilities provide structured access to PubMed and other Entrez databases; observe request
frequency guidance.
- EndNote documents `Reference Manager (RIS)` as an import option for RIS files.
- Nature Portfolio, AAAS, and Cell Press official pages should be checked when exact current journal
coverage matters.
references/ris-endnote.md›
# RIS, EndNote, and Zotero RDF Output
EndNote can import RIS files using the `Reference Manager (RIS)` import option. Use `.ris` as the
default exchange format because it is plain text, widely supported, and easy to inspect.
## Contents
- [RIS mapping for journal articles](#ris-mapping-for-journal-articles)
- [Author-integrity preflight](#author-integrity-preflight)
- [EndNote import instruction](#endnote-import-instruction)
- [Zotero RDF guidance](#zotero-rdf-guidance)
## RIS mapping for journal articles
Use these tags:
```text
TY - JOUR
TI - Article title
AU - Author, Given
AU - Consortium Name,
T2 - Journal title
JO - Journal title
PY - Publication year
Y1 - YYYY/MM/DD when available
VL - Volume
IS - Issue
SP - First page or article number
EP - Last page
DO - DOI
AN - PMID:12345678
UR - URL
SN - ISSN
N2 - Abstract or short metadata note, only when safely available
ER -
```
Rules:
- Write one `AU` line per author and preserve the source order.
- Personal names use `Family, Given` or `Family, Given, Suffix`. Keep source-provided initials and name particles.
- Corporate or consortium authors require a trailing comma, for example `AU - NMSS Validation Group,`. Double an embedded comma before adding the trailing comma.
- A surname-only personal line such as `AU - Chaudhuri` is incomplete. Do not deliver or import it as a finished record.
- Use `TY - JOUR` for journal articles.
- End every record with `ER -`.
- Do not invent missing fields.
- Write DOI in `DO`; a DOI URL in `UR` may be retained as well.
- Preserve a known PMID in `AN` so the record can be traced back to PubMed.
- Keep notes concise; avoid copying long abstracts into RIS unless the source terms allow it.
## Author-integrity preflight
Before export, verify all of the following:
1. The metadata source supplies a structured, ordered author list.
2. Every personal author has both a family name and a given name or initials.
3. Suffixes such as `Jr.` and `III`, name particles, and collective authors are retained.
4. One `AU` line is emitted for every source author. The exporter must not shorten the list to `et al.`.
5. Biomedical records with incomplete Crossref authors are refetched by PMID and compared against PubMed or the publisher page.
Bad:
```text
AU - Chaudhuri
AU - Schapira
```
Ready for EndNote:
```text
AU - Chaudhuri, K Ray
AU - Schapira, Anthony H V
```
The script blocks missing or surname-only author metadata by default. `--allow-incomplete-authors` is an explicit exception for cases where the limitation is understood; it must not be used to describe the export as verified.
## EndNote import instruction
Tell the user:
```text
In EndNote: File > Import > File, choose the `.ris` file, set Import Option to
Reference Manager (RIS), then import.
```
Menu labels vary slightly by EndNote version and operating system, so avoid over-specific UI claims
unless the user gives their exact EndNote version.
## Zotero RDF guidance
Use `.rdf` when the user explicitly asks for Zotero import/export.
Preferred structure:
```xml
<rdf:RDF ...>
<bib:Article rdf:about="https://doi.org/...">
<z:itemType>journalArticle</z:itemType>
<dcterms:isPartOf rdf:resource="urn:..."/>
<bib:authors>...</bib:authors>
<dc:title>...</dc:title>
<dc:date>YYYY-MM-DD</dc:date>
<dc:identifier>...</dc:identifier>
<bib:pages>...</bib:pages>
<z:citationKey>...</z:citationKey>
</bib:Article>
<bib:Journal rdf:about="urn:...">...</bib:Journal>
</rdf:RDF>
```
Rules:
- Export one `bib:Article` per citation.
- Represent authors as `foaf:Person` nodes inside `rdf:Seq`.
- Deduplicate journal container nodes by journal/ISSN/volume/issue identity.
- Do not invent abstracts, attachments, or fields that are not present in metadata.
references/script-usage.md›
# Script usage and long-article strategy
Open this reference when running `scripts/nature_citation.py` or when the input is long enough to need batching. It expands step 3 of the workflow.
## Running the script
Start with `scripts/nature_citation.py` when internet access is available:
```bash
python scripts/nature_citation.py \
--text "PASTE MANUSCRIPT TEXT HERE" \
--scope cns \
--outdir /tmp/nature-citation \
--format ris \
--with-artifacts
```
### Useful options
- `--text-file manuscript.txt`: read long text from a file.
- `--claim "CLAIM TEXT"` or `--claim-file claims.txt`: treat each claim as a segment.
- `--doi 10.xxxx/xxxxx` or `--doi-file dois.txt`: export known DOI records after screening.
- `--pmid 12345678` or `--pmid-file pmids.txt`: fetch the complete ordered author list and journal metadata directly from PubMed. Prefer this route for biomedical records when Crossref lacks given names or a consortium author.
- `--scope nature`: Nature Portfolio-style journals only.
- `--scope flagship`: Nature, Science, and Cell only.
- `--from-year 2018 --to-year 2026`: constrain publication dates.
- `--rows 40`: raise for broad searches; keep top candidates manageable.
- `--per-segment 3`: number of citation candidates to keep per segment.
- `--max-retries 2`: retry transient Crossref failures before skipping a query.
- `--format ris|enw|zotero-rdf`: export format. RIS is the default. If omitted and `--output-file` is set, infer from suffix.
- `--mailto [email protected]`: use Crossref's polite pool.
- `--ncbi-email [email protected]`: identify repeated PubMed E-utilities requests.
- `--allow-incomplete-authors`: bypass the default author-integrity block. Use only when an incomplete record is intentional; warning notes remain in the export.
- `--batch-size 10`: process segments in batches of N. Each batch writes an incremental export file.
- `--max-segments 20`: only process the first N segments. Useful for testing or section-by-section workflows.
- `--sleep 0.3`: seconds between Crossref requests. Default is 0.3; raise to 1.0 if rate-limited.
## Long-article strategy
When the input text is longer than roughly 3000 characters (about 10+ segments), switch to a batched workflow to avoid timeout, context overflow, or incomplete results:
1. **Auto-detect length.** Count segments after segmentation. If there are more than 10 segments, switch to batch mode automatically.
2. **Split by section.** Prefer splitting at paragraph double-line breaks or explicit section headings (`Introduction`, `Results`, etc.) so each batch is a coherent unit, not arbitrary sentence groups.
3. **Process each batch independently.** Run the script once per batch using `--batch-size` or `--max-segments`, OR split the text externally and call the script once per chunk. Each call writes its own intermediate export file.
4. **Merge results at the end.** After all batches finish, combine the intermediate files into one final export. Deduplicate by DOI.
5. **Minimize inline analysis.** For long articles, do NOT write detailed support-grade notes for every single segment inline. Instead:
- Write a compact summary table (segment ID → best candidate → support grade).
- Point the user to the HTML visualization for full browsing.
- Only elaborate on segments where no candidate was found or evidence is contradictory.
### Quick guide
| Segments | Strategy |
|---|---|
| 1–10 | Run once, full inline analysis is fine. |
| 11–25 | Use `--batch-size 10`. Write a compact summary table. Point to HTML. |
| 26+ | Split by section. Run script per section with `--batch-size 10`. Compact summary + HTML only. |
For long texts, prefer the HTML browser for review and selection instead of relying only on inline notes.
## Repairing incomplete author metadata
If the script reports an author-integrity failure, do not manually expand initials or guess names. Supply a PMID instead:
```bash
python scripts/nature_citation.py \
--pmid-file pmids.txt \
--format ris \
--output-file references.ris \
--ncbi-email [email protected]
```
The PubMed path retains personal given names or initials, suffixes, and collective authors. If PubMed is unavailable or the record is not indexed, verify the complete ordered author list against the publisher page before export.
references/search-strategy.md›
# Search Strategy
## Turn claims into searchable concepts
Break each sentence into:
- `phenomenon`: what is being claimed
- `entity`: gene, protein, pathway, compound, intervention, technology, population, or ecosystem
- `relationship`: increases, decreases, predicts, regulates, causes, associates with, improves, detects
- `context`: species, tissue, disease, cell type, geography, time period, device, method, or dataset
- `boundary`: "in cancer cells", "after treatment", "in older adults", "under drought", etc.
Create search queries at three levels:
1. `precise`: entity + relationship + outcome + context
2. `synonym`: alternate names and abbreviations
3. `broad`: field context if no direct paper is found
For Chinese claims, translate the scientific concepts, not the sentence literally. Keep acronyms and
standard nomenclature unchanged.
## Support grading
Use the smallest support grade that is defensible:
| Grade | Meaning | Good use |
|---|---|---|
| strong support | Directly tests the same core relationship in a similar context | Experimental, mechanistic, or quantitative manuscript claims |
| partial support | Supports one component or a narrower setting | Carefully qualified claims |
| background support | Establishes field context or prior observation | Introduction/background sentences |
| contradictory/limiting | Conflicts with or narrows the claim | Discussion, limitations, or avoid citing as support |
| metadata-only candidate | Metadata suggests relevance; abstract/full text not checked | Screening only |
## Evidence note template
```text
Claim: [original claim]
Paper: [first author/year/title/journal/DOI]
Support grade: [grade]
Evidence basis: [title/abstract/publisher page/full text]
Reasoning: [why the result supports or does not support the exact claim]
Citation wording: [how to phrase the manuscript sentence if using this citation]
```
## Common failure modes
- The paper is related to the same disease but tests a different mechanism.
- The paper supports an association, but the manuscript sentence claims causality.
- The evidence is in a different species, cell type, or clinical population.
- A review is used as primary evidence when original research exists.
- The claim is too broad for a single citation.
- The searched journal title contains "Nature" but is not a Nature Portfolio journal.
## Better search moves
- Add the method or model when results are broad: `single-cell`, `CRISPR screen`, `organoid`,
`randomized`, `cohort`, `meta-analysis`, `cryogenic electron microscopy`.
- Add context terms when there are many irrelevant hits: tissue, species, cell type, disease subtype,
exposure, intervention, or outcome.
- Search the opposite direction if the claim might be overconfident: `inhibits` vs `activates`,
`resistance` vs `sensitivity`, `risk` vs `protective`.
- Use recent limits for fast-moving areas, but remove them if no direct CNS/Nature-series paper appears.
scripts/nature_citation.py›
#!/usr/bin/env python3
"""
Segment manuscript text, search strict Nature/CNS-family citation candidates, and export a
reference-manager file. By default the script writes one inspectable `.ris` file.
Optional review artifacts can still be generated, but they are opt-in.
"""
from __future__ import annotations
import argparse
import csv
import hashlib
import html
import json
import re
import sys
import time
from dataclasses import dataclass, field
from pathlib import Path
from typing import Any, Callable
from urllib.parse import quote, urlencode
from urllib.request import Request, urlopen
import xml.etree.ElementTree as ET
from xml.sax.saxutils import escape as xml_escape
from xml.sax.saxutils import quoteattr
CROSSREF_API = "https://api.crossref.org/works"
PUBMED_EFETCH = "https://eutils.ncbi.nlm.nih.gov/entrez/eutils/efetch.fcgi"
USER_AGENT = "codex-nature-citation/1.0 (mailto:[email protected])"
EXPORT_FORMAT_CHOICES = ("enw", "ris", "zotero-rdf", "rdf")
DEFAULT_EXPORT_FORMAT = "ris"
ZOTERO_RDF_NS = {
"rdf": "http://www.w3.org/1999/02/22-rdf-syntax-ns#",
"z": "http://www.zotero.org/namespaces/export#",
"dcterms": "http://purl.org/dc/terms/",
"bib": "http://purl.org/net/biblio#",
"foaf": "http://xmlns.com/foaf/0.1/",
"dc": "http://purl.org/dc/elements/1.1/",
"prism": "http://prismstandard.org/namespaces/1.2/basic/",
}
NATURE_EXACT = {
"Nature",
"Nature Biotechnology",
"Nature Cancer",
"Nature Cardiovascular Research",
"Nature Cell Biology",
"Nature Chemical Biology",
"Nature Chemistry",
"Nature Climate Change",
"Nature Communications",
"Nature Computational Science",
"Nature Ecology & Evolution",
"Nature Electronics",
"Nature Energy",
"Nature Food",
"Nature Genetics",
"Nature Geoscience",
"Nature Human Behaviour",
"Nature Immunology",
"Nature Machine Intelligence",
"Nature Materials",
"Nature Medicine",
"Nature Metabolism",
"Nature Methods",
"Nature Microbiology",
"Nature Nanotechnology",
"Nature Neuroscience",
"Nature Photonics",
"Nature Physics",
"Nature Plants",
"Nature Protocols",
"Nature Reviews Cancer",
"Nature Reviews Cardiology",
"Nature Reviews Chemistry",
"Nature Reviews Clinical Oncology",
"Nature Reviews Drug Discovery",
"Nature Reviews Earth & Environment",
"Nature Reviews Endocrinology",
"Nature Reviews Gastroenterology & Hepatology",
"Nature Reviews Genetics",
"Nature Reviews Immunology",
"Nature Reviews Materials",
"Nature Reviews Microbiology",
"Nature Reviews Molecular Cell Biology",
"Nature Reviews Nephrology",
"Nature Reviews Neurology",
"Nature Reviews Neuroscience",
"Nature Reviews Physics",
"Nature Reviews Psychology",
"Nature Reviews Rheumatology",
"Nature Structural & Molecular Biology",
"Scientific Reports",
}
SCIENCE_EXACT = {
"Science",
"Science Advances",
"Science Immunology",
"Science Robotics",
"Science Signaling",
"Science Translational Medicine",
}
CELL_EXACT = {
"Cell",
"Cancer Cell",
"Cell Chemical Biology",
"Cell Genomics",
"Cell Host & Microbe",
"Cell Metabolism",
"Cell Reports",
"Cell Reports Medicine",
"Cell Reports Methods",
"Cell Reports Physical Science",
"Cell Stem Cell",
"Cell Systems",
"Chem",
"Current Biology",
"Developmental Cell",
"Immunity",
"Joule",
"Med",
"Molecular Cell",
"Neuron",
"One Earth",
"Patterns",
"Structure",
"The Innovation",
}
CELL_TRENDS_EXACT = {
"Trends in Biochemical Sciences",
"Trends in Biotechnology",
"Trends in Cancer",
"Trends in Cell Biology",
"Trends in Chemistry",
"Trends in Cognitive Sciences",
"Trends in Ecology & Evolution",
"Trends in Endocrinology & Metabolism",
"Trends in Genetics",
"Trends in Immunology",
"Trends in Microbiology",
"Trends in Molecular Medicine",
"Trends in Neurosciences",
"Trends in Parasitology",
"Trends in Pharmacological Sciences",
"Trends in Plant Science",
}
FLAGSHIP = {"Nature", "Science", "Cell"}
@dataclass
class Segment:
id: str
text: str
search_query: str
order: int
def as_dict(self) -> dict[str, Any]:
return {
"id": self.id,
"order": self.order,
"text": self.text,
"search_query": self.search_query,
}
@dataclass
class Candidate:
title: str
journal: str
family: str
year: str
y1: str
doi: str
url: str
volume: str
issue: str
start_page: str
end_page: str
issn: str
authors: list[str]
abstract: str
type: str
score: float
source_query: str
pmid: str = ""
source_catalog: str = "Crossref"
author_warnings: list[str] = field(default_factory=list)
@property
def doi_url(self) -> str:
return f"https://doi.org/{self.doi}" if self.doi else self.url
@property
def key(self) -> str:
if self.doi:
return self.doi.lower()
return f"{self.title.lower()}|{self.journal.lower()}"
@property
def first_author(self) -> str:
if not self.authors:
return "Unknown author"
return self.authors[0].split(",", 1)[0]
@property
def citation_marker(self) -> str:
if self.year:
return f"({self.first_author} et al., {self.year})"
return f"({self.first_author} et al.)"
@property
def page_range(self) -> str:
if self.start_page and self.end_page:
return f"{self.start_page}-{self.end_page}"
return self.start_page
@property
def identifier_url(self) -> str:
return self.doi_url or self.url
@property
def article_resource(self) -> str:
if self.identifier_url:
return self.identifier_url
return f"urn:candidate:{stable_hash(self.key or self.title or 'candidate')}"
@property
def journal_resource(self) -> str:
return build_journal_resource(self)
@property
def zotero_citation_key(self) -> str:
return build_zotero_citation_key(self)
def as_dict(self) -> dict[str, Any]:
return {
"title": self.title,
"journal": self.journal,
"family": self.family,
"year": self.year,
"doi": self.doi,
"pmid": self.pmid,
"url": self.url,
"doi_url": self.doi_url,
"volume": self.volume,
"issue": self.issue,
"start_page": self.start_page,
"end_page": self.end_page,
"issn": self.issn,
"authors": self.authors,
"abstract": self.abstract,
"type": self.type,
"score": self.score,
"source_query": self.source_query,
"source_catalog": self.source_catalog,
"author_metadata_status": "complete" if not self.author_warnings else "incomplete",
"author_warnings": self.author_warnings,
"citation_marker": self.citation_marker,
"support_grade": "metadata-only candidate",
"screening_note": "Inspect abstract/publisher page before citing this paper as support.",
"enw_record": build_enw_record(self),
"ris_record": build_ris_record(self),
"journal_resource": self.journal_resource,
"zotero_rdf_article": build_zotero_rdf_article(self),
"zotero_rdf_journal": build_zotero_rdf_journal(self),
}
def normalize_title(title: str) -> str:
return re.sub(r"\s+", " ", title or "").strip()
def stable_hash(value: str) -> str:
return hashlib.sha1(value.encode("utf-8")).hexdigest()[:12]
def slugify(value: str) -> str:
slug = re.sub(r"[^a-z0-9]+", "-", (value or "").lower()).strip("-")
return slug or "item"
def normalize_export_format(value: str | None) -> str:
if not value:
return DEFAULT_EXPORT_FORMAT
if value == "rdf":
return "zotero-rdf"
return value
def infer_export_format(output_path: Path | None) -> str:
if output_path is None:
return DEFAULT_EXPORT_FORMAT
suffix = output_path.suffix.lower()
if suffix == ".ris":
return "ris"
if suffix == ".rdf":
return "zotero-rdf"
if suffix == ".enw":
return "enw"
return DEFAULT_EXPORT_FORMAT
def export_filename(export_format: str, base: str = "references") -> str:
if export_format == "ris":
return f"{base}.ris"
if export_format == "zotero-rdf":
return f"{base}.rdf"
return f"{base}.enw"
def slug_from_text(text: str, max_words: int = 6) -> str:
"""Derive a filename slug from the first meaningful words of manuscript text."""
text = clean_text(text)
text = re.sub(r"\[[^\]]+\]|\([A-Za-z]+ et al\.,? \d{4}\)", " ", text)
words = re.findall(r"[A-Za-z0-9]+|[一-鿿]+", text)
stopwords = {
"the", "a", "an", "and", "or", "of", "to", "in", "for", "by", "with", "on", "at",
"from", "is", "are", "was", "were", "be", "been", "being", "that", "this", "these",
"those", "it", "its", "can", "may", "could", "not", "but", "as", "if", "into",
}
content = [w for w in words if w.lower() not in stopwords]
slug = "-".join(w.lower() for w in content[:max_words])
return slug or "references"
def export_label(export_format: str) -> str:
if export_format == "ris":
return "RIS"
if export_format == "zotero-rdf":
return "Zotero RDF"
return "ENW"
def make_partial_path(path: Path) -> Path:
return path.with_name(f"{path.stem}.partial{path.suffix}")
def retry_with_backoff(action: Callable[[], Any], max_retries: int, base_delay: float = 0.5) -> Any:
last_error: Exception | None = None
retries = max(0, max_retries)
for attempt in range(retries + 1):
try:
return action()
except Exception as exc: # noqa: BLE001
last_error = exc
if attempt >= retries:
break
time.sleep(base_delay * (2 ** attempt))
if last_error is not None:
raise last_error
raise RuntimeError("retry_with_backoff() exited without returning or raising")
def resolve_batch_size(segment_count: int, args: argparse.Namespace) -> int:
if getattr(args, "batch_size", 0) and args.batch_size > 0:
return max(1, args.batch_size)
if segment_count > 10:
return 10
return 0
def chunk_segments(segments: list[Segment], batch_size: int) -> list[list[Segment]]:
if not segments:
return []
if batch_size <= 0 or batch_size >= len(segments):
return [segments]
return [segments[idx : idx + batch_size] for idx in range(0, len(segments), batch_size)]
def limit_segments(segments: list[Segment], max_segments: int) -> tuple[list[Segment], int]:
if max_segments and max_segments > 0 and len(segments) > max_segments:
return segments[:max_segments], len(segments) - max_segments
return segments, 0
def zotero_date_value(item: Candidate) -> str:
if item.y1:
return item.y1.replace("/", "-")
return item.year
def split_author_parts(name: str) -> tuple[str, str]:
if name.endswith(","):
return name[:-1].replace(",,", ",").strip(), ""
if "," in name:
family, given = name.split(",", 1)
return family.strip(), given.strip()
parts = [part for part in name.split() if part]
if not parts:
return "", ""
if len(parts) == 1:
return parts[0], ""
return parts[-1], " ".join(parts[:-1])
def build_journal_resource(item: Candidate) -> str:
parts: list[str] = []
if item.issn:
parts.append(f"issn:{slugify(item.issn)}")
elif item.journal:
parts.append(f"title:{slugify(item.journal)}")
else:
parts.append(f"record:{stable_hash(item.key or item.title or 'journal')}")
if item.volume:
parts.append(f"vol:{slugify(item.volume)}")
if item.issue:
parts.append(f"issue:{slugify(item.issue)}")
return "urn:" + ":".join(parts)
def build_zotero_citation_key(item: Candidate) -> str:
first_author = slugify(item.first_author)
title_words = re.findall(r"[A-Za-z0-9]+", item.title)[:3]
title_part = "".join(word.capitalize() for word in title_words) or "Item"
year = item.year or "n.d."
return f"{first_author}{title_part}{year}"
def journal_family(journal: str) -> str | None:
journal = normalize_title(journal)
if not journal:
return None
if journal in NATURE_EXACT or journal.startswith("Nature ") or journal.startswith("npj "):
return "Nature Portfolio"
if journal in SCIENCE_EXACT:
return "Science family"
if journal in CELL_EXACT or journal in CELL_TRENDS_EXACT:
return "Cell Press"
return None
def in_scope(journal: str, scope: str) -> bool:
journal = normalize_title(journal)
if not journal:
return False
if scope == "flagship":
return journal in FLAGSHIP
family = journal_family(journal)
if scope == "nature":
return family == "Nature Portfolio"
if scope == "science":
return family == "Science family"
if scope == "cell":
return family == "Cell Press"
return family in {"Nature Portfolio", "Science family", "Cell Press"}
def first(values: list[Any] | None, default: str = "") -> str:
if not values:
return default
value = values[0]
if isinstance(value, str):
return value
return default
def date_parts(item: dict[str, Any]) -> list[int]:
for key in ("published-print", "published-online", "published", "issued"):
parts = item.get(key, {}).get("date-parts")
if parts and parts[0]:
return parts[0]
return []
def year_from_item(item: dict[str, Any]) -> str:
parts = date_parts(item)
return str(parts[0]) if parts else ""
def y1_from_item(item: dict[str, Any]) -> str:
parts = date_parts(item)
if not parts:
return ""
year = f"{parts[0]:04d}"
month = f"{parts[1]:02d}" if len(parts) > 1 else "01"
day = f"{parts[2]:02d}" if len(parts) > 2 else "01"
return f"{year}/{month}/{day}"
def endnote_corporate_author(name: str) -> str:
"""Return an EndNote-safe corporate author name.
EndNote requires a trailing comma to prevent a corporate name from being
inverted. Embedded commas are doubled so they remain part of the name.
"""
normalized = clean_text(name).rstrip(",").strip()
if not normalized:
return ""
return f"{normalized.replace(',', ',,')},"
def author_name(author: dict[str, Any]) -> str:
family = clean_text(str(author.get("family") or ""))
given = clean_text(str(author.get("given") or ""))
suffix = clean_text(str(author.get("suffix") or ""))
literal = clean_text(str(author.get("name") or ""))
if literal and not family and not given:
return endnote_corporate_author(literal)
if family and given:
parts = [family, given]
if suffix:
parts.append(suffix)
return ", ".join(parts)
return family or given or literal
def crossref_author_warnings(raw_authors: list[dict[str, Any]], authors: list[str]) -> list[str]:
warnings: list[str] = []
if not raw_authors or not authors:
return ["No structured author list was supplied by the metadata source."]
for index, author in enumerate(raw_authors, 1):
family = clean_text(str(author.get("family") or ""))
given = clean_text(str(author.get("given") or ""))
literal = clean_text(str(author.get("name") or ""))
if literal and not family and not given:
continue
if not family or not given:
warnings.append(
f"Author {index} lacks a {'family name' if not family else 'given name or initials'} in structured metadata."
)
return warnings
def pages(item: dict[str, Any]) -> tuple[str, str]:
page = item.get("page", "") or item.get("article-number", "")
if not page:
return "", ""
if "-" in page:
start, end = page.split("-", 1)
return start.strip(), end.strip()
return page.strip(), ""
def clean_text(text: str) -> str:
text = re.sub(r"<[^>]+>", " ", text or "")
text = re.sub(r"\s+", " ", text)
return text.strip()
def ris_escape(text: str) -> str:
return clean_text(text).replace("\n", " ").replace("\r", " ")
def split_sentences(text: str) -> list[str]:
text = re.sub(r"\s+", " ", text.strip())
if not text:
return []
pattern = r"(?<=[.!?。！？])\s+|(?<=[。！？])"
return [part.strip() for part in re.split(pattern, text) if part.strip()]
def looks_like_heading(text: str) -> bool:
stripped = text.strip()
if len(stripped) > 90:
return False
if stripped.endswith((".", "。", "!", "！", "?", "？")):
return False
words = stripped.split()
if 0 < len(words) <= 8 and not any(char in stripped for char in ",;，；"):
return True
return False
def query_from_segment(text: str, max_words: int = 26) -> str:
text = clean_text(text)
text = re.sub(r"\[[^\]]+\]|\([A-Za-z]+ et al\.,? \d{4}\)", " ", text)
words = re.findall(r"[A-Za-z0-9α-ωΑ-Ωβγδκλμνπρστυφχψω\-]+|[\u4e00-\u9fff]+", text)
if not words:
return text[:240]
return " ".join(words[:max_words])
def fallback_queries_from_segment(text: str) -> list[str]:
words = re.findall(r"[A-Za-z0-9α-ωΑ-Ωβγδκλμνπρστυφχψω\-]+|[\u4e00-\u9fff]+", clean_text(text))
if not words:
return []
stopwords = {
"the", "a", "an", "and", "or", "of", "to", "in", "for", "by", "with", "on", "at", "from",
"reveals", "reveal", "revealed", "promote", "promotes", "promoted", "promoting",
"suppress", "suppresses", "suppressing", "show", "shows", "showed", "indicate", "indicates",
"indicated", "identify", "identifies", "identified", "can", "may", "could", "is", "are",
"was", "were", "be", "been", "being", "that", "this", "these", "those",
}
content = [word for word in words if word.lower() not in stopwords]
candidates: list[str] = []
if len(content) >= 3:
candidates.append(" ".join(content[:12]))
if len(content) >= 5:
candidates.append(" ".join(content[:8]))
deduped: list[str] = []
seen: set[str] = set()
for candidate in candidates:
normalized = candidate.lower().strip()
if normalized and normalized not in seen:
seen.add(normalized)
deduped.append(candidate)
return deduped
def segment_text(text: str, max_chars: int = 700) -> list[Segment]:
normalized = text.replace("\r\n", "\n").replace("\r", "\n").strip()
if not normalized:
return []
paragraphs = [part.strip() for part in re.split(r"\n\s*\n+", normalized) if part.strip()]
raw_segments: list[str] = []
for paragraph in paragraphs:
if looks_like_heading(paragraph):
continue
sentences = split_sentences(paragraph)
if len(sentences) > 1:
raw_segments.extend(sentences)
elif len(paragraph) <= max_chars:
raw_segments.append(re.sub(r"\s+", " ", paragraph))
else:
raw_segments.extend(sentences)
segments: list[Segment] = []
for idx, segment in enumerate(raw_segments, 1):
cleaned = clean_text(segment)
if len(cleaned) < 10:
continue
segments.append(
Segment(
id=f"S{len(segments) + 1:03d}",
text=cleaned,
search_query=query_from_segment(cleaned),
order=len(segments) + 1,
)
)
return segments
def candidate_from_crossref(item: dict[str, Any], source_query: str) -> Candidate | None:
journal = first(item.get("container-title"))
if not journal:
return None
family = journal_family(journal) or ""
start, end = pages(item)
raw_authors = item.get("author", [])
authors = [author_name(author) for author in raw_authors]
authors = [author for author in authors if author]
return Candidate(
title=clean_text(first(item.get("title"))),
journal=normalize_title(journal),
family=family,
year=year_from_item(item),
y1=y1_from_item(item),
doi=item.get("DOI", ""),
url=item.get("URL", ""),
volume=item.get("volume", ""),
issue=item.get("issue", ""),
start_page=start,
end_page=end,
issn=first(item.get("ISSN")),
authors=authors,
abstract=clean_text(item.get("abstract", "")),
type=item.get("type", ""),
score=float(item.get("score", 0.0) or 0.0),
source_query=source_query,
source_catalog="Crossref",
author_warnings=crossref_author_warnings(raw_authors, authors),
)
def crossref_headers(mailto: str | None = None) -> dict[str, str]:
return {"User-Agent": USER_AGENT if not mailto else f"codex-nature-citation/1.0 (mailto:{mailto})"}
def fetch_crossref(query: str, rows: int, mailto: str | None = None, from_year: int | None = None, to_year: int | None = None, retries: int = 2) -> list[dict[str, Any]]:
filters = ["type:journal-article"]
if from_year is not None:
filters.append(f"from-pub-date:{from_year}-01-01")
if to_year is not None:
filters.append(f"until-pub-date:{to_year}-12-31")
params = {
"query.bibliographic": query,
"rows": str(rows),
"select": "DOI,title,container-title,published,published-print,published-online,issued,author,volume,issue,page,article-number,ISSN,URL,abstract,type,score",
"filter": ",".join(filters),
"sort": "relevance",
"order": "desc",
}
if mailto:
params["mailto"] = mailto
url = f"{CROSSREF_API}?{urlencode(params)}"
req = Request(url, headers=crossref_headers(mailto))
last_exc: Exception | None = None
for attempt in range(1, retries + 2):
try:
with urlopen(req, timeout=30) as response:
payload = json.loads(response.read().decode("utf-8"))
return payload.get("message", {}).get("items", [])
except Exception as exc: # noqa: BLE001
last_exc = exc
if attempt <= retries:
time.sleep(min(2 ** attempt, 8))
raise last_exc # type: ignore[misc]
def fetch_crossref_doi(doi: str, mailto: str | None = None) -> dict[str, Any]:
url = f"{CROSSREF_API}/{quote(doi.strip(), safe='')}"
if mailto:
url = f"{url}?{urlencode({'mailto': mailto})}"
req = Request(url, headers=crossref_headers(mailto))
with urlopen(req, timeout=30) as response:
payload = json.loads(response.read().decode("utf-8"))
return payload.get("message", {})
MONTHS = {
"jan": 1,
"feb": 2,
"mar": 3,
"apr": 4,
"may": 5,
"jun": 6,
"jul": 7,
"aug": 8,
"sep": 9,
"oct": 10,
"nov": 11,
"dec": 12,
}
def xml_text(element: ET.Element | None) -> str:
if element is None:
return ""
return clean_text("".join(element.itertext()))
def pubmed_date(article: ET.Element) -> tuple[str, str]:
date_node = article.find("MedlineCitation/Article/Journal/JournalIssue/PubDate")
if date_node is None:
date_node = article.find("MedlineCitation/Article/ArticleDate")
year = xml_text(date_node.find("Year")) if date_node is not None else ""
if not year and date_node is not None:
match = re.search(r"\b(18|19|20|21)\d{2}\b", xml_text(date_node.find("MedlineDate")))
year = match.group(0) if match else ""
if not year:
return "", ""
month_text = xml_text(date_node.find("Month")) if date_node is not None else ""
if month_text.isdigit():
month = max(1, min(12, int(month_text)))
else:
month = MONTHS.get(month_text[:3].lower(), 1)
day_text = xml_text(date_node.find("Day")) if date_node is not None else ""
day = max(1, min(31, int(day_text))) if day_text.isdigit() else 1
return year, f"{int(year):04d}/{month:02d}/{day:02d}"
def split_page_text(page: str) -> tuple[str, str]:
page = clean_text(page)
if not page:
return "", ""
if "-" in page:
start, end = page.split("-", 1)
return start.strip(), end.strip()
return page, ""
def pubmed_authors(article: ET.Element) -> tuple[list[str], list[str]]:
authors: list[str] = []
warnings: list[str] = []
author_nodes = article.findall("MedlineCitation/Article/AuthorList/Author")
if not author_nodes:
return [], ["No structured author list was supplied by PubMed."]
for index, author in enumerate(author_nodes, 1):
collective = xml_text(author.find("CollectiveName"))
if collective:
authors.append(endnote_corporate_author(collective))
continue
family = xml_text(author.find("LastName"))
given = xml_text(author.find("ForeName")) or xml_text(author.find("Initials"))
suffix = xml_text(author.find("Suffix"))
if family and given:
name_parts = [family, given]
if suffix:
name_parts.append(suffix)
authors.append(", ".join(name_parts))
continue
incomplete = family or given
if incomplete:
authors.append(incomplete)
warnings.append(
f"Author {index} lacks a {'family name' if not family else 'given name or initials'} in PubMed metadata."
)
if not authors and not warnings:
warnings.append("No usable authors were found in PubMed metadata.")
return authors, warnings
def candidate_from_pubmed(article: ET.Element) -> Candidate | None:
article_node = article.find("MedlineCitation/Article")
if article_node is None:
return None
pmid = xml_text(article.find("MedlineCitation/PMID"))
journal = xml_text(article_node.find("Journal/Title")) or xml_text(
article.find("MedlineCitation/MedlineJournalInfo/MedlineTA")
)
title = xml_text(article_node.find("ArticleTitle"))
if not title or not journal:
return None
year, y1 = pubmed_date(article)
start_page, end_page = split_page_text(xml_text(article_node.find("Pagination/MedlinePgn")))
doi = ""
for identifier in article.findall("PubmedData/ArticleIdList/ArticleId"):
if identifier.attrib.get("IdType", "").lower() == "doi":
doi = xml_text(identifier)
break
if not doi:
for identifier in article_node.findall("ELocationID"):
if identifier.attrib.get("EIdType", "").lower() == "doi":
doi = xml_text(identifier)
break
authors, warnings = pubmed_authors(article)
abstract = " ".join(
xml_text(node) for node in article_node.findall("Abstract/AbstractText") if xml_text(node)
)
return Candidate(
title=title,
journal=normalize_title(journal),
family=journal_family(journal) or "",
year=year,
y1=y1,
doi=doi,
url=f"https://pubmed.ncbi.nlm.nih.gov/{pmid}/" if pmid else "",
volume=xml_text(article_node.find("Journal/JournalIssue/Volume")),
issue=xml_text(article_node.find("Journal/JournalIssue/Issue")),
start_page=start_page,
end_page=end_page,
issn=xml_text(article_node.find("Journal/ISSN")),
authors=authors,
abstract=abstract,
type=xml_text(article_node.find("PublicationTypeList/PublicationType")) or "journal-article",
score=0.0,
source_query=f"pmid:{pmid}",
pmid=pmid,
source_catalog="PubMed",
author_warnings=warnings,
)
def fetch_pubmed(pmids: list[str], email: str | None = None) -> list[ET.Element]:
params = {
"db": "pubmed",
"id": ",".join(pmids),
"retmode": "xml",
"tool": "nature-citation",
}
if email:
params["email"] = email
request = Request(
f"{PUBMED_EFETCH}?{urlencode(params)}",
headers={"User-Agent": USER_AGENT if not email else f"codex-nature-citation/1.0 (mailto:{email})"},
)
with urlopen(request, timeout=45) as response:
root = ET.fromstring(response.read())
return list(root.findall("PubmedArticle"))
def dedupe(candidates: list[Candidate]) -> list[Candidate]:
output: list[Candidate] = []
positions: dict[str, int] = {}
for candidate in candidates:
if not candidate.key:
continue
if candidate.key not in positions:
positions[candidate.key] = len(output)
output.append(candidate)
continue
existing_index = positions[candidate.key]
existing = output[existing_index]
existing_quality = (not existing.author_warnings, len(existing.authors), bool(existing.doi))
candidate_quality = (not candidate.author_warnings, len(candidate.authors), bool(candidate.doi))
if candidate_quality > existing_quality:
output[existing_index] = candidate
return output
def author_integrity_issues(candidates: list[Candidate]) -> list[str]:
issues: list[str] = []
for candidate in candidates:
identifier = candidate.doi or (f"PMID {candidate.pmid}" if candidate.pmid else candidate.title)
candidate_issues = list(candidate.author_warnings)
if not candidate.authors and not candidate_issues:
candidate_issues.append("No authors are present.")
for index, author in enumerate(candidate.authors, 1):
if "," not in author and not any(issue.startswith(f"Author {index} ") for issue in candidate_issues):
candidate_issues.append(
f"Author {index} is not encoded as a complete personal name or an EndNote corporate author."
)
for issue in candidate_issues:
issues.append(f"{identifier}: {issue}")
return issues
def validate_author_integrity(candidates: list[Candidate], allow_incomplete: bool = False) -> None:
issues = author_integrity_issues(candidates)
if not issues or allow_incomplete:
return
preview = "\n".join(f" - {issue}" for issue in issues[:12])
remainder = f"\n - ... and {len(issues) - 12} more issue(s)" if len(issues) > 12 else ""
raise ValueError(
"Author metadata integrity check failed. Export was stopped so surname-only, missing, or "
"otherwise incomplete author data is not presented as EndNote-ready. Refetch the affected "
"records with --pmid/--pmid-file or verify them against the publisher record. Use "
"--allow-incomplete-authors only when the limitation is intentional.\n"
f"{preview}{remainder}"
)
def build_ris_record(item: Candidate) -> str:
lines: list[str] = []
lines.append("TY - JOUR")
if item.title:
lines.append(f"TI - {ris_escape(item.title)}")
for author in item.authors:
lines.append(f"AU - {ris_escape(author)}")
if item.journal:
lines.append(f"T2 - {ris_escape(item.journal)}")
lines.append(f"JO - {ris_escape(item.journal)}")
if item.year:
lines.append(f"PY - {ris_escape(item.year)}")
if item.y1:
lines.append(f"Y1 - {ris_escape(item.y1)}")
if item.volume:
lines.append(f"VL - {ris_escape(item.volume)}")
if item.issue:
lines.append(f"IS - {ris_escape(item.issue)}")
if item.start_page:
lines.append(f"SP - {ris_escape(item.start_page)}")
if item.end_page:
lines.append(f"EP - {ris_escape(item.end_page)}")
if item.doi:
lines.append(f"DO - {ris_escape(item.doi)}")
if item.pmid:
lines.append(f"AN - PMID:{ris_escape(item.pmid)}")
if item.doi_url:
lines.append(f"UR - {ris_escape(item.doi_url)}")
if item.issn:
lines.append(f"SN - {ris_escape(item.issn)}")
for warning in item.author_warnings:
lines.append(f"N1 - Author metadata warning: {ris_escape(warning)}")
lines.append("N1 - Metadata-only candidate. Inspect abstract or publisher page before citing as support.")
lines.append("ER -")
return "\n".join(lines)
def write_ris(candidates: list[Candidate], path: Path, allow_incomplete_authors: bool = False) -> None:
validate_author_integrity(candidates, allow_incomplete=allow_incomplete_authors)
lines: list[str] = []
for item in candidates:
lines.append(build_ris_record(item))
lines.append("")
path.write_text("\n".join(lines), encoding="utf-8")
def build_enw_record(item: Candidate) -> str:
lines: list[str] = []
lines.append("%0 Journal Article")
if item.title:
lines.append(f"%T {ris_escape(item.title)}")
for author in item.authors:
lines.append(f"%A {ris_escape(author)}")
if item.journal:
lines.append(f"%J {ris_escape(item.journal)}")
if item.volume:
lines.append(f"%V {ris_escape(item.volume)}")
if item.issue:
lines.append(f"%N {ris_escape(item.issue)}")
if item.start_page and item.end_page:
lines.append(f"%P {ris_escape(item.start_page)}-{ris_escape(item.end_page)}")
elif item.start_page:
lines.append(f"%P {ris_escape(item.start_page)}")
if item.year:
lines.append(f"%D {ris_escape(item.year)}")
if item.issn:
lines.append(f"%@ {ris_escape(item.issn)}")
if item.doi:
lines.append(f"%R {ris_escape(item.doi)}")
if item.pmid:
lines.append(f"%M PMID:{ris_escape(item.pmid)}")
if item.doi_url:
lines.append(f"%U {ris_escape(item.doi_url)}")
return "\n".join(lines)
def write_enw(candidates: list[Candidate], path: Path, allow_incomplete_authors: bool = False) -> None:
validate_author_integrity(candidates, allow_incomplete=allow_incomplete_authors)
lines: list[str] = []
for item in candidates:
lines.append(build_enw_record(item))
lines.append("")
path.write_text("\n".join(lines), encoding="utf-8")
def build_zotero_rdf_article(item: Candidate) -> str:
lines: list[str] = [f' <bib:Article rdf:about={quoteattr(item.article_resource)}>']
lines.append(" <z:itemType>journalArticle</z:itemType>")
if item.journal:
lines.append(f' <dcterms:isPartOf rdf:resource={quoteattr(item.journal_resource)}/>')
if item.authors:
lines.append(" <bib:authors>")
lines.append(" <rdf:Seq>")
for author in item.authors:
family, given = split_author_parts(author)
lines.append(" <rdf:li>")
lines.append(" <foaf:Person>")
if family:
lines.append(f" <foaf:surname>{xml_escape(family)}</foaf:surname>")
if given:
lines.append(f" <foaf:givenName>{xml_escape(given)}</foaf:givenName>")
lines.append(" </foaf:Person>")
lines.append(" </rdf:li>")
lines.append(" </rdf:Seq>")
lines.append(" </bib:authors>")
if item.title:
lines.append(f" <dc:title>{xml_escape(item.title)}</dc:title>")
date_value = zotero_date_value(item)
if date_value:
lines.append(f" <dc:date>{xml_escape(date_value)}</dc:date>")
lines.append(f" <z:libraryCatalog>{xml_escape(item.source_catalog)}</z:libraryCatalog>")
if item.identifier_url:
lines.append(" <dc:identifier>")
lines.append(" <dcterms:URI>")
lines.append(f" <rdf:value>{xml_escape(item.identifier_url)}</rdf:value>")
lines.append(" </dcterms:URI>")
lines.append(" </dc:identifier>")
if item.doi:
lines.append(f" <dc:identifier>{xml_escape(f'DOI {item.doi}')}</dc:identifier>")
if item.page_range:
lines.append(f" <bib:pages>{xml_escape(item.page_range)}</bib:pages>")
lines.append(f" <z:citationKey>{xml_escape(item.zotero_citation_key)}</z:citationKey>")
lines.append(" </bib:Article>")
return "\n".join(lines)
def build_zotero_rdf_journal(item: Candidate) -> str:
lines: list[str] = [f' <bib:Journal rdf:about={quoteattr(item.journal_resource)}>']
if item.volume:
lines.append(f" <prism:volume>{xml_escape(item.volume)}</prism:volume>")
if item.journal:
lines.append(f" <dc:title>{xml_escape(item.journal)}</dc:title>")
if item.issue:
lines.append(f" <prism:number>{xml_escape(item.issue)}</prism:number>")
if item.issn:
lines.append(f" <dc:identifier>{xml_escape(f'ISSN {item.issn}')}</dc:identifier>")
lines.append(" </bib:Journal>")
return "\n".join(lines)
def build_zotero_rdf_document(candidates: list[Candidate]) -> str:
root_open = [
"<rdf:RDF",
*(f' xmlns:{prefix}="{uri}"' for prefix, uri in ZOTERO_RDF_NS.items()),
">",
]
journal_map: dict[str, str] = {}
article_blocks: list[str] = []
for item in candidates:
article_blocks.append(build_zotero_rdf_article(item))
if item.journal and item.journal_resource not in journal_map:
journal_map[item.journal_resource] = build_zotero_rdf_journal(item)
sections = ["".join(root_open), *article_blocks, *journal_map.values(), "</rdf:RDF>"]
return "\n".join(section for section in sections if section)
def write_zotero_rdf(candidates: list[Candidate], path: Path, allow_incomplete_authors: bool = False) -> None:
validate_author_integrity(candidates, allow_incomplete=allow_incomplete_authors)
path.write_text(build_zotero_rdf_document(candidates), encoding="utf-8")
def read_text_inputs(args: argparse.Namespace) -> str:
parts: list[str] = []
if args.text:
parts.extend(args.text)
if args.text_file:
parts.append(Path(args.text_file).read_text(encoding="utf-8"))
return "\n\n".join(part for part in parts if part.strip())
def read_claims(args: argparse.Namespace) -> list[str]:
claims: list[str] = []
if args.claim:
claims.extend(args.claim)
if args.claim_file:
for line in Path(args.claim_file).read_text(encoding="utf-8").splitlines():
line = line.strip()
if line:
claims.append(line)
return claims
def read_dois(args: argparse.Namespace) -> list[str]:
dois: list[str] = []
if args.doi:
dois.extend(args.doi)
if args.doi_file:
for line in Path(args.doi_file).read_text(encoding="utf-8").splitlines():
line = line.strip()
if line and not line.startswith("#"):
dois.append(line)
cleaned = []
for doi in dois:
doi = doi.strip()
doi = re.sub(r"^https?://(?:dx\.)?doi\.org/", "", doi, flags=re.IGNORECASE)
if doi:
cleaned.append(doi)
return cleaned
def read_pmids(args: argparse.Namespace) -> list[str]:
raw_pmids: list[str] = []
if args.pmid:
raw_pmids.extend(args.pmid)
if args.pmid_file:
for line in Path(args.pmid_file).read_text(encoding="utf-8").splitlines():
line = line.strip()
if line and not line.startswith("#"):
raw_pmids.append(line)
pmids: list[str] = []
seen: set[str] = set()
for value in raw_pmids:
normalized = re.sub(r"^PMID\s*:\s*", "", value.strip(), flags=re.IGNORECASE)
normalized = re.sub(r"^https?://pubmed\.ncbi\.nlm\.nih\.gov/", "", normalized, flags=re.IGNORECASE)
normalized = normalized.strip("/ ")
if not re.fullmatch(r"\d+", normalized):
raise ValueError(f"Invalid PMID: {value}")
if normalized not in seen:
seen.add(normalized)
pmids.append(normalized)
return pmids
def build_segments(args: argparse.Namespace) -> list[Segment]:
text = read_text_inputs(args)
segments = segment_text(text, max_chars=args.segment_chars) if text else []
claims = read_claims(args)
for claim in claims:
cleaned = clean_text(claim)
if cleaned:
segments.append(
Segment(
id=f"S{len(segments) + 1:03d}",
text=cleaned,
search_query=query_from_segment(cleaned),
order=len(segments) + 1,
)
)
return segments
def search_segment(segment: Segment, args: argparse.Namespace) -> tuple[list[Candidate], list[dict[str, str]]]:
errors: list[dict[str, str]] = []
candidates: list[Candidate] = []
queries = [segment.search_query, *fallback_queries_from_segment(segment.text)]
seen_queries: set[str] = set()
for query in queries:
normalized_query = query.strip().lower()
if not normalized_query or normalized_query in seen_queries:
continue
seen_queries.add(normalized_query)
try:
items = retry_with_backoff(
lambda: fetch_crossref(
query,
rows=args.rows,
mailto=args.mailto,
from_year=args.from_year,
to_year=args.to_year,
),
max_retries=args.max_retries,
)
except Exception as exc: # noqa: BLE001
errors.append({"segment_id": segment.id, "query": query, "error": str(exc)})
continue
for item in items:
candidate = candidate_from_crossref(item, source_query=query)
if candidate and in_scope(candidate.journal, args.scope):
candidates.append(candidate)
if dedupe(candidates):
break
return dedupe(candidates)[: args.per_segment], errors
def build_mapping(segments: list[Segment], args: argparse.Namespace) -> tuple[list[dict[str, Any]], list[Candidate], list[dict[str, str]]]:
mapping: list[dict[str, Any]] = []
all_candidates: list[Candidate] = []
errors: list[dict[str, str]] = []
for segment in segments:
candidates, segment_errors = search_segment(segment, args)
errors.extend(segment_errors)
all_candidates.extend(candidates)
mapping.append(
{
"segment": segment,
"references": candidates,
"suggested_insert_text": " ".join(candidate.citation_marker for candidate in candidates[: args.citations_per_segment]),
}
)
if args.sleep:
time.sleep(args.sleep)
return mapping, dedupe(all_candidates), errors
def summarize_mapping(mapping: list[dict[str, Any]], references: list[Candidate], errors: list[dict[str, str]]) -> str:
return (
f"segments={len(mapping)} "
f"references={len(references)} "
f"errors={len(errors)}"
)
def write_export_checkpoint(
outdir: Path,
base_path: Path,
export_format: str,
references: list[Candidate],
allow_incomplete_authors: bool = False,
) -> Path:
partial_output = make_partial_path(base_path)
if export_format == "enw":
write_enw(references, partial_output, allow_incomplete_authors=allow_incomplete_authors)
elif export_format == "ris":
write_ris(references, partial_output, allow_incomplete_authors=allow_incomplete_authors)
else:
write_zotero_rdf(references, partial_output, allow_incomplete_authors=allow_incomplete_authors)
return partial_output
def write_final_artifacts(
mapping: list[dict[str, Any]],
references: list[Candidate],
outdir: Path,
output_path: Path,
args: argparse.Namespace,
errors: list[dict[str, str]],
skipped_segments: int = 0,
) -> tuple[Path, Path, Path, Path]:
artifact_base = outdir / (output_path.stem if output_path.stem else "citation")
json_payload = mapping_to_json(mapping, references, args, errors)
if skipped_segments:
json_payload["notes"].append(f"Skipped {skipped_segments} segment(s) because --max-segments was set.")
json_path = artifact_base.with_suffix(".json")
tsv_path = artifact_base.with_suffix(".tsv")
report_path = artifact_base.with_suffix(".md")
html_path = artifact_base.with_suffix(".html")
json_path.write_text(json.dumps(json_payload, ensure_ascii=False, indent=2), encoding="utf-8")
write_mapping_tsv(mapping, tsv_path)
write_report(mapping, report_path, args.scope, len(references), args.format, output_path.name)
write_html(mapping, references, outdir, html_path, output_path, args.format)
return json_path, tsv_path, report_path, html_path
def process_segment_batches(
segments: list[Segment],
args: argparse.Namespace,
outdir: Path,
base_path: Path,
) -> tuple[list[dict[str, Any]], list[Candidate], list[dict[str, str]]]:
batch_size = resolve_batch_size(len(segments), args)
batches = chunk_segments(segments, batch_size)
mapping: list[dict[str, Any]] = []
references: list[Candidate] = []
errors: list[dict[str, str]] = []
if not batches:
return mapping, references, errors
for batch_index, batch in enumerate(batches, 1):
print(
f"Processing batch {batch_index}/{len(batches)}: segments {batch[0].order}-{batch[-1].order} ({len(batch)} segments)..."
)
batch_mapping, batch_references, batch_errors = build_mapping(batch, args)
mapping.extend(batch_mapping)
references = dedupe([*references, *batch_references])
errors.extend(batch_errors)
try:
partial_output = write_export_checkpoint(
outdir,
base_path,
args.format,
references,
allow_incomplete_authors=args.allow_incomplete_authors,
)
except ValueError as exc:
partial_output = None
print(f" Checkpoint not written: {str(exc).splitlines()[0]}", file=sys.stderr)
print(
f" Batch {batch_index} done: {sum(len(entry['references']) for entry in batch_mapping)} candidates, "
f"cumulative {len(references)} unique refs."
)
if partial_output:
print(f" Checkpoint saved: {partial_output}")
return mapping, references, errors
def fetch_doi_candidates(dois: list[str], args: argparse.Namespace) -> tuple[list[Candidate], list[dict[str, str]]]:
candidates: list[Candidate] = []
errors: list[dict[str, str]] = []
for doi in dois:
try:
item = retry_with_backoff(
lambda: fetch_crossref_doi(doi, mailto=args.mailto),
max_retries=args.max_retries,
)
except Exception as exc: # noqa: BLE001
errors.append({"doi": doi, "error": str(exc)})
continue
candidate = candidate_from_crossref(item, source_query=f"doi:{doi}")
if candidate:
candidates.append(candidate)
if args.sleep:
time.sleep(args.sleep)
return dedupe(candidates), errors
def fetch_pubmed_candidates(pmids: list[str], args: argparse.Namespace) -> tuple[list[Candidate], list[dict[str, str]]]:
candidates: list[Candidate] = []
errors: list[dict[str, str]] = []
for offset in range(0, len(pmids), 100):
batch = pmids[offset : offset + 100]
try:
articles = retry_with_backoff(
lambda: fetch_pubmed(batch, email=args.ncbi_email or args.mailto),
max_retries=args.max_retries,
)
except Exception as exc: # noqa: BLE001
errors.append({"pmids": ",".join(batch), "error": str(exc)})
continue
returned: set[str] = set()
for article in articles:
candidate = candidate_from_pubmed(article)
if candidate:
candidates.append(candidate)
if candidate.pmid:
returned.add(candidate.pmid)
for missing in set(batch) - returned:
errors.append({"pmid": missing, "error": "PubMed returned no usable journal-article metadata."})
if args.sleep and offset + 100 < len(pmids):
time.sleep(args.sleep)
return dedupe(candidates), errors
def mapping_to_json(mapping: list[dict[str, Any]], references: list[Candidate], args: argparse.Namespace, errors: list[dict[str, str]]) -> dict[str, Any]:
return {
"scope": args.scope,
"from_year": args.from_year,
"to_year": args.to_year,
"export_format": args.format,
"segment_count": len(mapping),
"reference_count": len(references),
"segments": [
{
**entry["segment"].as_dict(),
"suggested_insert_text": entry["suggested_insert_text"],
"references": [candidate.as_dict() for candidate in entry["references"]],
}
for entry in mapping
],
"references": [candidate.as_dict() for candidate in references],
"errors": errors,
"notes": [
"Candidates are retrieved from Crossref and/or PubMed metadata and require abstract/full-text screening before citation.",
"Author metadata passes a strict integrity gate unless --allow-incomplete-authors is explicitly set.",
"Supported reference-manager exports: ENW, RIS, and Zotero RDF.",
],
}
def write_mapping_tsv(mapping: list[dict[str, Any]], path: Path) -> None:
fields = [
"segment_id",
"segment_order",
"segment_text",
"search_query",
"suggested_insert_text",
"citation_marker",
"support_grade",
"title",
"journal",
"family",
"year",
"doi",
"pmid",
"doi_url",
"authors",
"score",
"screening_note",
]
with path.open("w", newline="", encoding="utf-8") as fh:
writer = csv.DictWriter(fh, fieldnames=fields, delimiter="\t")
writer.writeheader()
for entry in mapping:
segment: Segment = entry["segment"]
if not entry["references"]:
writer.writerow(
{
"segment_id": segment.id,
"segment_order": segment.order,
"segment_text": segment.text,
"search_query": segment.search_query,
"suggested_insert_text": "",
"support_grade": "no candidate",
"screening_note": "No in-scope candidate found in Crossref metadata search.",
}
)
for candidate in entry["references"]:
writer.writerow(
{
"segment_id": segment.id,
"segment_order": segment.order,
"segment_text": segment.text,
"search_query": segment.search_query,
"suggested_insert_text": entry["suggested_insert_text"],
"citation_marker": candidate.citation_marker,
"support_grade": "metadata-only candidate",
"title": candidate.title,
"journal": candidate.journal,
"family": candidate.family,
"year": candidate.year,
"doi": candidate.doi,
"pmid": candidate.pmid,
"doi_url": candidate.doi_url,
"authors": "; ".join(candidate.authors[:10]),
"score": candidate.score,
"screening_note": "Inspect abstract/publisher page before citing this paper as support.",
}
)
def write_report(
mapping: list[dict[str, Any]],
path: Path,
scope: str,
reference_count: int,
export_format: str,
output_name: str,
) -> None:
lines = [
"# Nature Citation Report",
"",
"## Search Scope",
"",
f"- Scope: `{scope}`",
f"- Segments: {len(mapping)}",
f"- Unique references exported: {reference_count}",
"- Sources: Crossref metadata search and optional PubMed PMID retrieval",
"- Author integrity: incomplete or surname-only structured names are blocked by default.",
"- Support grade: all rows are `metadata-only candidate` until abstract/full text is checked.",
"",
"## Segment-to-Reference Map",
"",
]
for entry in mapping:
segment: Segment = entry["segment"]
lines.extend([f"### {segment.id}", "", segment.text, ""])
if entry["suggested_insert_text"]:
lines.append(f"Suggested insert text: `{entry['suggested_insert_text']}`")
lines.append("")
if not entry["references"]:
lines.extend(["- No in-scope candidate found.", ""])
continue
for candidate in entry["references"]:
lines.extend(
[
f"- {candidate.citation_marker} {candidate.title}. *{candidate.journal}* ({candidate.year}). {candidate.doi_url}",
f" - Family: {candidate.family or 'Unclassified'}",
" - Support grade: metadata-only candidate",
]
)
lines.append("")
lines.extend(
[
"## Export",
"",
f"- Output file: `{output_name}`",
f"- Format: `{export_format}` ({export_label(export_format)})",
"- RIS is the default inspectable exchange format and imports into EndNote with Reference Manager (RIS).",
"- ENW is suitable for EndNote tagged import when explicitly selected.",
"- Zotero RDF is suitable for direct Zotero import.",
]
)
path.write_text("\n".join(lines), encoding="utf-8")
def rel_link(path: Path, outdir: Path) -> str:
try:
return html.escape(path.relative_to(outdir).as_posix())
except ValueError:
return html.escape(path.as_posix())
def write_html(
mapping: list[dict[str, Any]],
references: list[Candidate],
outdir: Path,
path: Path,
export_path: Path,
export_format: str,
) -> None:
payload = {
"segmentCount": len(mapping),
"referenceCount": len(references),
"defaultExportFormat": export_format,
"defaultExportName": export_path.name,
"references": [candidate.as_dict() for candidate in references],
"segments": [
{
"id": entry["segment"].id,
"order": entry["segment"].order,
"text": entry["segment"].text,
"query": entry["segment"].search_query,
"suggestedInsertText": entry["suggested_insert_text"],
"references": [candidate.as_dict() for candidate in entry["references"]],
}
for entry in mapping
],
}
payload_json = json.dumps(payload, ensure_ascii=False).replace("</script>", "<\\/script>")
cards: list[str] = []
for entry in mapping:
segment: Segment = entry["segment"]
refs = entry["references"]
ref_items = []
if refs:
for idx, candidate in enumerate(refs, 1):
record_json = html.escape(json.dumps(candidate.as_dict(), ensure_ascii=False))
ref_items.append(
f"""
<article
class="reference"
data-key="{html.escape(candidate.key)}"
data-title="{html.escape(candidate.title)}"
data-journal="{html.escape(candidate.journal)}"
data-doi="{html.escape(candidate.doi)}"
data-year="{html.escape(candidate.year)}"
data-authors="{html.escape('; '.join(candidate.authors[:10]))}"
data-segment="{html.escape(segment.text)}"
data-record="{record_json}"
>
<div class="ref-controls">
<label class="check-row"><input type="checkbox"> Select</label>
<span class="marker">{html.escape(candidate.citation_marker)}</span>
</div>
<div class="ref-topline">
<span class="grade">metadata-only</span>
<span>{idx} / {len(refs)}</span>
</div>
<h3>{html.escape(candidate.title)}</h3>
<p class="meta">{html.escape(candidate.journal)} · {html.escape(candidate.family or "Unclassified")} · {html.escape(candidate.year or "n.d.")}</p>
<p class="authors">{html.escape("; ".join(candidate.authors[:6]))}</p>
<p class="abstract">{html.escape(candidate.abstract[:380] + ('...' if len(candidate.abstract) > 380 else '')) if candidate.abstract else 'No abstract snippet available from Crossref metadata.'}</p>
<a href="{html.escape(candidate.doi_url)}" target="_blank" rel="noreferrer">{html.escape(candidate.doi_url)}</a>
</article>
"""
)
else:
ref_items.append('<p class="empty">No in-scope candidate was found for this segment.</p>')
cards.append(
f"""
<section class="segment-card" id="{html.escape(segment.id)}">
<div class="segment-head">
<span class="segment-id">{html.escape(segment.id)} · Sentence {segment.order}</span>
<span class="query">Query: {html.escape(segment.search_query)}</span>
</div>
<p class="segment-text">{html.escape(segment.text)}</p>
<div class="insert-row">
<span>Suggested insert</span>
<code>{html.escape(entry["suggested_insert_text"] or "No candidate")}</code>
</div>
<div class="references-grid">
{"".join(ref_items)}
</div>
</section>
"""
)
export_link = rel_link(export_path, outdir)
export_label_text = html.escape(export_label(export_format))
export_file_label = html.escape(export_path.name)
doc = f"""<!doctype html>
<html lang="en">
<head>
<meta charset="utf-8">
<meta name="viewport" content="width=device-width, initial-scale=1">
<title>Nature Citation Browser</title>
<style>
:root {{
--ink: #1f1f1f;
--muted: #5f6368;
--paper: #f8f9fa;
--panel: #ffffff;
--line: #dadce0;
--accent: #1a73e8;
--accent-2: #d93025;
--soft: #e8f0fe;
--soft-2: #f1f3f4;
}}
* {{ box-sizing: border-box; }}
body {{
margin: 0;
color: var(--ink);
background: var(--paper);
font-family: Arial, Helvetica, sans-serif;
line-height: 1.45;
}}
header {{
padding: 26px clamp(16px, 4vw, 40px) 16px;
border-bottom: 1px solid var(--line);
background: var(--panel);
}}
h1 {{
margin: 0 0 8px;
font-size: clamp(28px, 3vw, 42px);
font-weight: 500;
letter-spacing: 0;
}}
.subhead {{
max-width: 980px;
margin: 0;
color: var(--muted);
font-size: 15px;
}}
.toolbar {{
display: grid;
grid-template-columns: 260px minmax(0, 1fr) auto;
gap: 14px;
align-items: start;
padding: 16px clamp(16px, 4vw, 40px);
background: var(--panel);
border-bottom: 1px solid var(--line);
position: sticky;
top: 0;
z-index: 8;
}}
.toolbar a, .toolbar button, .toolbar input, .toolbar select {{
appearance: none;
border: 1px solid var(--line);
background: var(--panel);
color: var(--ink);
border-radius: 20px;
padding: 10px 14px;
font: 500 14px Arial, Helvetica, sans-serif;
text-decoration: none;
}}
.toolbar button, .toolbar a {{
cursor: pointer;
}}
.toolbar button.primary, .toolbar a.primary {{
background: var(--accent);
border-color: var(--accent);
color: white;
}}
.filters {{
display: grid;
gap: 10px;
}}
.filter-group {{
display: flex;
gap: 10px;
flex-wrap: wrap;
align-items: center;
}}
.search-input {{
width: 100%;
}}
.summary-strip {{
display: flex;
gap: 10px;
flex-wrap: wrap;
align-items: center;
justify-content: center;
}}
.pill {{
border: 1px solid var(--line);
border-radius: 999px;
padding: 9px 12px;
background: var(--soft-2);
font-size: 13px;
}}
.layout {{
display: grid;
grid-template-columns: minmax(0, 1fr) 320px;
gap: 18px;
padding: 22px clamp(16px, 4vw, 40px) 48px;
}}
main {{
display: grid;
gap: 14px;
align-content: start;
}}
aside {{
position: sticky;
top: 124px;
align-self: start;
background: var(--panel);
border: 1px solid var(--line);
border-radius: 12px;
padding: 16px;
box-shadow: 0 2px 8px rgba(60, 64, 67, 0.12);
}}
.segment-card {{
background: var(--panel);
border: 1px solid var(--line);
border-radius: 12px;
padding: 18px 20px;
box-shadow: 0 1px 4px rgba(60, 64, 67, 0.12);
}}
.segment-head {{
display: flex;
gap: 12px;
align-items: start;
justify-content: space-between;
flex-wrap: wrap;
margin-bottom: 10px;
}}
.segment-id {{
background: var(--soft);
color: var(--accent);
border-radius: 999px;
padding: 5px 10px;
font-size: 12px;
font-weight: 700;
}}
.query {{
color: var(--muted);
font-size: 12px;
max-width: 780px;
}}
.segment-text {{
margin: 0 0 14px;
font-size: 18px;
}}
.insert-row {{
display: grid;
gap: 6px;
margin: 12px 0 16px;
padding: 12px;
background: var(--soft);
border-radius: 10px;
}}
.insert-row span {{
color: var(--muted);
font-size: 12px;
font-weight: 700;
}}
code {{
white-space: normal;
font-family: ui-monospace, SFMono-Regular, Menlo, Consolas, monospace;
font-size: 14px;
}}
.references-grid {{
display: grid;
gap: 12px;
}}
.reference {{
border: 1px solid var(--line);
border-radius: 12px;
padding: 14px 16px;
background: #fff;
}}
.ref-topline {{
display: flex;
justify-content: space-between;
gap: 10px;
margin-bottom: 8px;
font-size: 12px;
}}
.marker {{
color: var(--accent);
font-weight: 800;
}}
.grade {{
color: var(--accent-2);
font-weight: 800;
}}
.ref-controls {{
display: flex;
justify-content: space-between;
align-items: center;
gap: 12px;
margin-bottom: 10px;
}}
.check-row {{
display: inline-flex;
align-items: center;
gap: 8px;
font-size: 13px;
color: var(--muted);
}}
h3 {{
margin: 0 0 8px;
font-size: 19px;
line-height: 1.35;
font-weight: 500;
}}
.meta, .authors, .empty {{
margin: 0 0 8px;
color: var(--muted);
font-size: 14px;
}}
.reference a {{
color: var(--accent);
overflow-wrap: anywhere;
font-size: 14px;
font-weight: 500;
}}
.abstract {{
margin: 8px 0 0;
color: #3c4043;
font-size: 14px;
}}
.hidden {{
display: none !important;
}}
.modal-overlay {{
position: fixed;
inset: 0;
display: flex;
align-items: center;
justify-content: center;
padding: 14px;
background: rgba(20, 33, 31, 0.42);
backdrop-filter: blur(3px);
z-index: 20;
}}
.modal-overlay.is-hidden {{
display: none;
}}
.modal-panel {{
border: 1px solid var(--ink);
border-radius: 8px;
width: min(680px, calc(100vw - 28px));
box-shadow: 0 16px 40px rgba(60, 64, 67, 0.28);
background: var(--panel);
color: var(--ink);
}}
.modal-inner {{
padding: 24px;
}}
.modal-inner h2 {{
margin: 0 0 10px;
font-size: 28px;
}}
.stats {{
display: grid;
grid-template-columns: repeat(2, minmax(0, 1fr));
gap: 10px;
margin: 18px 0;
}}
.stat {{
border: 1px solid var(--line);
border-radius: 10px;
padding: 14px;
background: var(--soft-2);
}}
.stat b {{
display: block;
font-size: 30px;
line-height: 1;
}}
.modal-actions {{
display: flex;
flex-wrap: wrap;
gap: 10px;
margin-top: 16px;
}}
.modal-actions a, .modal-actions button {{
border: 1px solid var(--accent);
background: var(--accent);
color: #fffdf8;
border-radius: 20px;
padding: 10px 12px;
font: 700 14px Arial, Helvetica, sans-serif;
text-decoration: none;
cursor: pointer;
}}
.modal-actions button {{
background: var(--panel);
color: var(--accent);
}}
.selection-title {{
margin: 0 0 10px;
font-size: 16px;
font-weight: 700;
}}
.selection-meta {{
margin: 0 0 14px;
color: var(--muted);
font-size: 13px;
}}
.selection-list {{
display: grid;
gap: 10px;
margin-bottom: 14px;
max-height: 48vh;
overflow: auto;
}}
.selection-item {{
border: 1px solid var(--line);
border-radius: 10px;
padding: 10px;
background: var(--soft-2);
}}
.selection-item h4 {{
margin: 0 0 6px;
font-size: 14px;
font-weight: 600;
}}
.selection-item p {{
margin: 0;
color: var(--muted);
font-size: 12px;
}}
.empty-state {{
color: var(--muted);
font-size: 14px;
padding: 12px;
border: 1px dashed var(--line);
border-radius: 10px;
background: var(--soft-2);
}}
@media (max-width: 680px) {{
.stats {{ grid-template-columns: 1fr; }}
.segment-head {{ align-items: flex-start; }}
}}
@media (max-width: 1040px) {{
.toolbar {{
grid-template-columns: 1fr;
}}
.summary-strip {{
justify-content: flex-start;
}}
.layout {{
grid-template-columns: 1fr;
}}
aside {{
position: static;
}}
}}
</style>
</head>
<body>
<header>
<h1>Nature Citation Browser</h1>
<p class="subhead">Scholar-style browsing for strict Nature/CNS-family candidates. Filter by year, compare related hits, choose the references you actually want, then download them as RIS, ENW, or Zotero RDF.</p>
</header>
<nav class="toolbar">
<div class="filters">
<input id="queryFilter" class="search-input" type="text" placeholder="Filter by title, journal, DOI, author, or segment text">
<div class="filter-group">
<label for="yearFrom">From year</label>
<input id="yearFrom" type="number" min="1900" max="2100" placeholder="2019">
<label for="yearTo">To year</label>
<input id="yearTo" type="number" min="1900" max="2100" placeholder="2026">
<button id="applyFilters" type="button">Apply filters</button>
<button id="clearFilters" type="button">Clear</button>
</div>
<div class="filter-group">
<label class="check-row"><input id="selectedOnly" type="checkbox"> Show selected only</label>
</div>
</div>
<div class="summary-strip">
<span class="pill"><strong id="visibleCount">0</strong> visible</span>
<span class="pill"><strong id="selectedCount">0</strong> selected</span>
<span class="pill"><strong>{payload["segmentCount"]}</strong> segments</span>
</div>
<div class="summary-strip">
<button id="openSummary" type="button">Guide</button>
<label for="downloadFormat">Format</label>
<select id="downloadFormat">
<option value="ris">RIS</option>
<option value="enw">ENW</option>
<option value="zotero-rdf">Zotero RDF</option>
</select>
<a class="primary" id="downloadSelected" href="{export_link}" download="{export_file_label}">Download selected / all</a>
</div>
</nav>
<div class="layout">
<main id="results">
{"".join(cards)}
</main>
<aside>
<h2 class="selection-title">Selected references</h2>
<p class="selection-meta">Tick the references you want. The download button exports the selection in the chosen format. If nothing is selected, it exports all references.</p>
<div id="selectionList" class="selection-list"></div>
<div id="selectionEmpty" class="empty-state">No references selected yet.</div>
</aside>
</div>
<div class="modal-overlay" id="summaryModal" aria-hidden="false">
<div class="modal-panel" role="dialog" aria-modal="true" aria-labelledby="summaryTitle">
<div class="modal-inner">
<h2 id="summaryTitle">Ready to review citations</h2>
<p>This page behaves more like a compact Scholar browser: filter by year, browse related results, select the citations you trust, and download them as ENW, RIS, or Zotero RDF.</p>
<div class="stats">
<div class="stat"><b>{payload["segmentCount"]}</b><span>segments</span></div>
<div class="stat"><b>{payload["referenceCount"]}</b><span>unique references</span></div>
</div>
<p class="meta">Important: these are still metadata-level candidates. Selection is your decision, not an automatic validity guarantee.</p>
<div class="modal-actions">
<a href="{export_link}" download>Download original {export_label_text} file</a>
<button id="closeSummary" type="button" onclick="document.getElementById('summaryModal').classList.add('is-hidden'); document.getElementById('summaryModal').setAttribute('aria-hidden', 'true');">Start browsing</button>
</div>
</div>
</div>
</div>
<script id="citationData" type="application/json">{payload_json}</script>
<script>
const data = JSON.parse(document.getElementById('citationData').textContent);
const modal = document.getElementById('summaryModal');
const closeSummary = document.getElementById('closeSummary');
const openSummary = document.getElementById('openSummary');
const yearFromInput = document.getElementById('yearFrom');
const yearToInput = document.getElementById('yearTo');
const queryFilterInput = document.getElementById('queryFilter');
const selectedOnlyInput = document.getElementById('selectedOnly');
const applyFiltersButton = document.getElementById('applyFilters');
const clearFiltersButton = document.getElementById('clearFilters');
const visibleCount = document.getElementById('visibleCount');
const selectedCount = document.getElementById('selectedCount');
const selectionList = document.getElementById('selectionList');
const selectionEmpty = document.getElementById('selectionEmpty');
const downloadSelected = document.getElementById('downloadSelected');
const downloadFormat = document.getElementById('downloadFormat');
const selectedMap = new Map();
let currentDownloadUrl = null;
downloadFormat.value = data.defaultExportFormat || 'ris';
function showModal() {{
if (modal) {{
modal.classList.remove('is-hidden');
modal.setAttribute('aria-hidden', 'false');
}}
}}
function hideModal() {{
if (modal) {{
modal.classList.add('is-hidden');
modal.setAttribute('aria-hidden', 'true');
}}
}}
window.addEventListener('DOMContentLoaded', showModal);
openSummary.addEventListener('click', showModal);
closeSummary.addEventListener('click', hideModal);
modal.addEventListener('click', (event) => {{
if (event.target === modal) {{
hideModal();
}}
}});
function revokeDownloadUrl() {{
if (currentDownloadUrl) {{
URL.revokeObjectURL(currentDownloadUrl);
currentDownloadUrl = null;
}}
}}
function selectedItems() {{
return Array.from(selectedMap.values());
}}
function exportItems() {{
const items = selectedItems();
return items.length ? items : (data.references || []);
}}
function buildEnw(items) {{
return items.map((item) => item.enw_record).join('\\n\\n');
}}
function buildRis(items) {{
return items.map((item) => item.ris_record).join('\\n\\n');
}}
function buildZoteroRdf(items) {{
const journals = new Map();
const articles = [];
for (const item of items) {{
if (item.zotero_rdf_article) {{
articles.push(item.zotero_rdf_article);
}}
if (item.journal_resource && item.zotero_rdf_journal && !journals.has(item.journal_resource)) {{
journals.set(item.journal_resource, item.zotero_rdf_journal);
}}
}}
const header = [
'<rdf:RDF',
' xmlns:rdf="{ZOTERO_RDF_NS["rdf"]}"',
' xmlns:z="{ZOTERO_RDF_NS["z"]}"',
' xmlns:dcterms="{ZOTERO_RDF_NS["dcterms"]}"',
' xmlns:bib="{ZOTERO_RDF_NS["bib"]}"',
' xmlns:foaf="{ZOTERO_RDF_NS["foaf"]}"',
' xmlns:dc="{ZOTERO_RDF_NS["dc"]}"',
' xmlns:prism="{ZOTERO_RDF_NS["prism"]}"',
'>',
].join('');
return [header, ...articles, ...journals.values(), '</rdf:RDF>'].join('\\n');
}}
function exportFileName(format, hasSelection) {{
const base = hasSelection ? 'selected-references' : 'all-references';
if (format === 'ris') return `${{base}}.ris`;
if (format === 'zotero-rdf') return `${{base}}.rdf`;
return `${{base}}.enw`;
}}
function updateDownloadLink() {{
const items = exportItems();
const format = downloadFormat.value || 'enw';
const hasSelection = selectedItems().length > 0;
let content = '';
if (format === 'ris') {{
content = buildRis(items);
}} else if (format === 'zotero-rdf') {{
content = buildZoteroRdf(items);
}} else {{
content = buildEnw(items);
}}
revokeDownloadUrl();
currentDownloadUrl = URL.createObjectURL(new Blob([content], {{ type: 'text/plain;charset=utf-8' }}));
downloadSelected.href = currentDownloadUrl;
downloadSelected.download = exportFileName(format, hasSelection);
}}
function updateSelectedPanel() {{
selectionList.innerHTML = '';
const items = selectedItems();
selectedCount.textContent = String(items.length);
if (!items.length) {{
selectionEmpty.classList.remove('hidden');
}} else {{
selectionEmpty.classList.add('hidden');
}}
for (const item of items) {{
const wrapper = document.createElement('div');
wrapper.className = 'selection-item';
wrapper.innerHTML = `<h4>${{item.title}}</h4><p>${{item.journal}} · ${{item.year || 'n.d.'}}</p>`;
selectionList.appendChild(wrapper);
}}
updateDownloadLink();
}}
function yearPass(year) {{
const fromYear = Number(yearFromInput.value || 0);
const toYear = Number(yearToInput.value || 0);
const y = Number(year || 0);
if (fromYear && (!y || y < fromYear)) return false;
if (toYear && (!y || y > toYear)) return false;
return true;
}}
function queryPass(text) {{
const needle = queryFilterInput.value.trim().toLowerCase();
if (!needle) return true;
return text.toLowerCase().includes(needle);
}}
function applyFilters() {{
let totalVisible = 0;
document.querySelectorAll('.segment-card').forEach((card) => {{
let visibleRefs = 0;
card.querySelectorAll('.reference').forEach((ref) => {{
const title = ref.dataset.title || '';
const journal = ref.dataset.journal || '';
const doi = ref.dataset.doi || '';
const authors = ref.dataset.authors || '';
const segmentText = ref.dataset.segment || '';
const year = ref.dataset.year || '';
const key = ref.dataset.key || '';
const searchText = `${{title}} ${{journal}} ${{doi}} ${{authors}} ${{segmentText}}`;
const matches = yearPass(year) && queryPass(searchText) && (!selectedOnlyInput.checked || selectedMap.has(key));
ref.classList.toggle('hidden', !matches);
if (matches) visibleRefs += 1;
}});
const noRefsMessage = card.querySelector('.empty');
if (noRefsMessage) {{
const matches = !selectedOnlyInput.checked && queryPass(card.dataset.segment || '');
noRefsMessage.classList.toggle('hidden', !matches);
card.classList.toggle('hidden', !matches);
if (matches) totalVisible += 1;
return;
}}
card.classList.toggle('hidden', visibleRefs === 0);
totalVisible += visibleRefs;
}});
visibleCount.textContent = String(totalVisible);
}}
document.querySelectorAll('.reference').forEach((ref) => {{
const checkbox = ref.querySelector('input[type="checkbox"]');
checkbox.addEventListener('change', () => {{
const key = ref.dataset.key;
const record = JSON.parse(ref.dataset.record);
if (checkbox.checked) {{
selectedMap.set(key, record);
}} else {{
selectedMap.delete(key);
}}
updateSelectedPanel();
applyFilters();
}});
}});
applyFiltersButton.addEventListener('click', applyFilters);
downloadFormat.addEventListener('change', updateDownloadLink);
clearFiltersButton.addEventListener('click', () => {{
yearFromInput.value = '';
yearToInput.value = '';
queryFilterInput.value = '';
selectedOnlyInput.checked = false;
applyFilters();
}});
queryFilterInput.addEventListener('input', applyFilters);
selectedOnlyInput.addEventListener('change', applyFilters);
yearFromInput.addEventListener('change', applyFilters);
yearToInput.addEventListener('change', applyFilters);
updateSelectedPanel();
applyFilters();
</script>
</body>
</html>
"""
path.write_text(doc, encoding="utf-8")
def parse_args(argv: list[str]) -> argparse.Namespace:
parser = argparse.ArgumentParser(description="Segment text and export strict Nature/CNS-family citations for EndNote or Zotero.")
parser.add_argument("--text", action="append", help="Manuscript text to segment and cite. Can be repeated.")
parser.add_argument("--text-file", help="UTF-8 manuscript text file.")
parser.add_argument("--claim", action="append", help="Single claim to treat as one segment. Can be repeated.")
parser.add_argument("--claim-file", help="UTF-8 text file with one claim per line.")
parser.add_argument("--doi", action="append", help="Known DOI to fetch and export. Can be repeated.")
parser.add_argument("--doi-file", help="UTF-8 text file with one DOI per line.")
parser.add_argument("--pmid", action="append", help="Known PubMed PMID to fetch and export. Can be repeated.")
parser.add_argument("--pmid-file", help="UTF-8 text file with one PMID per line.")
parser.add_argument("--scope", choices=["cns", "nature", "science", "cell", "flagship"], default="cns")
parser.add_argument("--output-file", help="Reference output file path, typically ending in .enw, .ris, or .rdf.")
parser.add_argument("--outdir", help="Optional directory for outputs. If omitted, uses the output file parent or current directory.")
parser.add_argument("--format", choices=EXPORT_FORMAT_CHOICES, help="Reference export format: enw, ris, or zotero-rdf. Inferred from --output-file when omitted.")
parser.add_argument("--with-artifacts", action="store_true", help="Also generate JSON/TSV/report/HTML review artifacts.")
parser.add_argument("--rows", type=int, default=30, help="Crossref rows per segment before journal-scope filtering.")
parser.add_argument("--per-segment", type=int, default=3, help="Maximum candidates to keep per segment.")
parser.add_argument("--citations-per-segment", type=int, default=2, help="Markers to include in suggested insert text.")
parser.add_argument("--segment-chars", type=int, default=700, help="Split paragraphs longer than this many characters.")
parser.add_argument("--max-candidates", type=int, default=80, help="Maximum deduplicated references to export.")
parser.add_argument("--max-segments", type=int, help="Limit the number of segments processed in a single run.")
parser.add_argument("--batch-size", type=int, help="Process segments in batches of this size.")
parser.add_argument("--max-retries", type=int, default=2, help="Maximum retry count for Crossref requests.")
parser.add_argument("--from-year", type=int, help="Earliest publication year.")
parser.add_argument("--to-year", type=int, help="Latest publication year.")
parser.add_argument("--mailto", help="Email for Crossref polite pool.")
parser.add_argument("--ncbi-email", help="Contact email sent to NCBI E-utilities for PMID requests.")
parser.add_argument(
"--allow-incomplete-authors",
action="store_true",
help="Export despite missing/surname-only author metadata. Disabled by default to protect EndNote imports.",
)
parser.add_argument("--sleep", type=float, default=0.3, help="Seconds between Crossref requests.")
return parser.parse_args(argv)
def main(argv: list[str]) -> int:
args = parse_args(argv)
segments = build_segments(args)
dois = read_dois(args)
try:
pmids = read_pmids(args)
except ValueError as exc:
print(str(exc), file=sys.stderr)
return 2
if not segments and not dois and not pmids:
print(
"Provide --text, --text-file, --claim, --claim-file, --doi, --doi-file, --pmid, or --pmid-file.",
file=sys.stderr,
)
return 2
output_path = Path(args.output_file).expanduser().resolve() if args.output_file else None
if args.outdir:
outdir = Path(args.outdir).expanduser().resolve()
elif output_path:
outdir = output_path.parent
else:
outdir = Path.cwd().resolve()
outdir.mkdir(parents=True, exist_ok=True)
args.format = normalize_export_format(args.format) if args.format else infer_export_format(output_path)
# Derive a meaningful base name from input text when no explicit output file was given
raw_text = read_text_inputs(args)
name_base = slug_from_text(raw_text) if not args.output_file else None
if output_path is None:
output_path = outdir / export_filename(args.format, base=name_base or "references")
segments, skipped_segments = limit_segments(segments, args.max_segments or 0)
mapping, references, errors = process_segment_batches(segments, args, outdir, output_path)
doi_candidates, doi_errors = fetch_doi_candidates(dois, args)
errors.extend(doi_errors)
pubmed_candidates, pubmed_errors = fetch_pubmed_candidates(pmids, args)
errors.extend(pubmed_errors)
references = dedupe([*references, *doi_candidates, *pubmed_candidates])[: args.max_candidates]
if (dois or pmids) and not references:
print("No requested DOI/PMID records could be retrieved; no export was written.", file=sys.stderr)
for error in errors[:5]:
print(f" - {error}", file=sys.stderr)
return 4
try:
if args.format == "enw":
write_enw(references, output_path, allow_incomplete_authors=args.allow_incomplete_authors)
elif args.format == "ris":
write_ris(references, output_path, allow_incomplete_authors=args.allow_incomplete_authors)
else:
write_zotero_rdf(references, output_path, allow_incomplete_authors=args.allow_incomplete_authors)
except ValueError as exc:
print(str(exc), file=sys.stderr)
return 3
if args.with_artifacts:
artifact_base = outdir / name_base if name_base else outdir / "citation"
json_path, tsv_path, report_path, html_path = write_final_artifacts(
mapping,
references,
outdir,
output_path,
args,
errors,
skipped_segments=skipped_segments,
)
print(f"Artifacts: {html_path}, {tsv_path}, {json_path}, {report_path}")
print(f"Reference output: {output_path}")
print(f"Export format: {args.format} ({export_label(args.format)})")
print(f"Unique references exported: {len(references)}")
if skipped_segments:
print(f"Segments skipped: {skipped_segments}")
if errors:
detail = "; see the JSON artifact for details" if args.with_artifacts else ""
print(f"Encountered {len(errors)} retrieval error(s){detail}.", file=sys.stderr)
return 0
if __name__ == "__main__":
raise SystemExit(main(sys.argv[1:]))
SKILL.md›
---
name: nature-citation
description: >-
Add strict Nature/CNS citations to manuscript text by splitting long passages into citable
segments, searching only accepted flagship and subjournal titles from Nature Portfolio, the
AAAS Science family, and Cell Press, filtering by publication time range, and exporting one
reference-manager-ready output by default. Use this skill whenever the user asks to input text and
automatically get references, add citations to a paragraph/manuscript, find Nature-series or CNS
support for statements, create text-to-reference correspondence, "分段引用", "自动给出引用",
"Nature系列引用", "CNS及子刊", "支撑文献", "补引用", "找引用", or export EndNote/RIS/ENW/Zotero RDF.
Also trigger on general academic-writing citation needs even without the word "Nature", such as
adding references while writing a paper, finding sources/literature for a claim, building a
reference list, citation/referencing for academic writing, and Chinese phrasings like
学术写作引用、写论文加引用、写paper找文献、加参考文献、配文献、引用文献、文献支撑.
metadata:
author: Yuan1z skill, refactored into static/dynamic layers
---
# Nature Citation — Router
This skill is split into two layers:
- A **static layer** under `static/` that holds versioned, reusable content fragments (core principles and scope, the Chinese-user operating mode, and the citation workflow).
- A **dynamic layer** (this file plus `manifest.yaml`) that loads the core every time and reaches for heavier material only when a step needs it.
Do not try to apply the citation logic from memory or from this router. Always load fragments from disk as described below.
## Routing protocol
Follow these four steps every time the skill is invoked.
### 1. Load the manifest and the core layer
Read [manifest.yaml](manifest.yaml). Then read every file listed under `always_load`:
- `static/core/principles.md` — what the skill produces, the strict journal scope, the source hierarchy, and the search-quality rules.
- `static/core/chinese-mode.md` — how to operate when the user writes in Chinese or asks for `Nature系列`/`CNS及子刊` style support.
- `static/core/workflow.md` — the seven-step workflow and the final report format.
### 2. No content axis — confirm scope and language inline
Unlike the other nature-* skills, nature-citation has no fragment axis. Its variation is runtime parameters, not different content bodies:
- **journal scope** — `Nature系列` / `CNS` / `CNS及子刊` / flagship-only. Read it from the user's wording (see `core/principles.md`) and pass it to the script as `--scope`.
- **user language** — if the user writes Chinese, follow `core/chinese-mode.md` (Chinese notes, English search queries).
- **input length** — if there are more than ~10 segments, switch to the batched long-article strategy in `references/script-usage.md`.
State the detected scope and date limits in one short line before searching.
### 3. Run the workflow
Follow the seven steps in `core/workflow.md`: segment, parse, search, evaluate support conservatively, validate complete structured author metadata, export one reference-manager file, generate review artifacts when useful, and report with the HTML browser path first. Prefer `scripts/nature_citation.py` for the search/export when internet access is available; open `references/script-usage.md` for its full flag list and the long-article batch strategy. When DOI metadata lacks given names, refetch the record by PMID or verify it against the publisher rather than exporting surname-only `AU` fields.
Never present a paper as support merely because its title is related, and never cite a metadata-only candidate without checking the abstract or publisher page. Do not invent missing bibliographic fields.
### 4. Reach for references only when needed
The files under `references/` are deep references, not defaults. Open them on demand per the `references.on_demand` table in the manifest:
- running the script, full flags, long-article batching → `references/script-usage.md`.
- turning a claim into search queries and support grades → `references/search-strategy.md`.
- the exact Nature/CNS journal-family boundary → `references/journal-scope.md`.
- RIS / EndNote / Zotero RDF export details → `references/ris-endnote.md`.
## Why this split
- The static layer is versioned and reviewable; the core stays small for a normal short run.
- The dynamic layer keeps each invocation cheap: the script flag dump and long-article strategy load only when actually running a search.
- The router itself is short on purpose. Update fragments and references, not this file, when adding scope.
- This structure mirrors `nature-writing`, `nature-polishing`, `nature-reader`, `nature-paper2ppt`, and `nature-figure`.
static/core/chinese-mode.md›
# Chinese-user operating mode
When the user writes in Chinese, asks for "Nature系列", "CNS及其子刊", "支撑文献", "补引用", "自动给出引用", "分段引用", "导出EndNote", "RIS", "Zotero", "RDF", or provides Chinese manuscript text:
- Accept the text in Chinese, but search using English concept queries unless the topic is explicitly China-specific or Chinese-language scholarship.
- Return segment notes and evidence notes in Chinese by default.
- Preserve the exact source segment and translate it into one or more English search claims.
- Flag overclaiming clearly in Chinese: `强支撑`, `部分支撑`, `背景支撑`, `不建议引用为该句支撑`.
- Do not present a paper as supporting the claim merely because its title is related.
static/core/principles.md›
# Core principles (citation)
Use this skill to turn manuscript text into a defensible citation export:
- segmented text with citation candidates for each segment
- a reference-manager import file in `.enw`, `.ris`, or Zotero `.rdf`
- conservative evidence notes explaining whether each candidate truly supports the segment
## Default scope
Interpret journal scope from the user's wording, but keep the filter strict:
- `Nature系列`: search Nature Portfolio first. Include `Nature`, `Nature [field]`, `Nature Communications`, `Communications [field]`, `Scientific Reports`, and `npj` journals.
- `CNS`: search `Cell`, `Nature`, and `Science` plus their major sister journals.
- `CNS及其子刊` or `CNS/sister journals`: search only accepted flagship and subjournal titles in Nature Portfolio, the AAAS Science family, and Cell Press.
- `只要Nature/Science/Cell正刊`: restrict to the flagship journals `Nature`, `Science`, and `Cell`.
Do not treat merely related journals as in-scope. A title is valid only if it is in the accepted publisher-family whitelist or clearly matches the official naming pattern for that family. If the user needs an exhaustive or submission-critical boundary, verify current official journal pages before finalizing because journal portfolios change. The exact boundary and official source notes are in `references/journal-scope.md`.
## Source hierarchy
Use sources in this order:
1. Structured bibliographic metadata: Crossref, PubMed/NCBI E-utilities, DOI metadata.
2. Publisher pages: `nature.com`, `science.org`, `cell.com`, and official journal pages.
3. Full text or abstract pages, if accessible.
4. Secondary databases such as Google Scholar, Semantic Scholar, Web of Science, or Scopus only as discovery aids, not as the sole support basis.
Prefer structured APIs for metadata and publisher pages for claim verification. If metadata and publisher page disagree, preserve the DOI and journal-page facts and flag the discrepancy.
## Search quality rules
- Prefer precision over volume. A useful answer is usually 3-8 candidates, not 50 loosely related papers.
- Use exact phrase searches only for distinctive terms; otherwise use concept terms and synonyms.
- Check journal identity. Many journals contain the word "nature" but are not Nature Portfolio journals.
- Treat citation count as a tie-breaker, not evidence of support.
- Capture retractions, corrections, and expressions of concern when visible in Crossref or publisher metadata.
- Date-sensitive topics require current searching and an explicit search date.
- For medical, clinical, or safety claims, search current literature and state that citations do not replace clinical guidance or systematic review.
## Source notes
This skill is based on public bibliographic APIs and official publisher/import documentation: Crossref REST API and filters, NCBI E-utilities, EndNote RIS import options, Nature Portfolio, AAAS Science journals, and Cell Press portfolio descriptions. Verify pages at use time when exact journal coverage or current import behavior matters.
static/core/workflow.md›
# Workflow
Run these seven steps for any citation job. For runs with more than ~10 segments, switch to the batched long-article strategy in `references/script-usage.md`.
## 1. Segment the text
- Split long text into citable segments. Prefer paragraph boundaries first, then sentence boundaries.
- Keep each segment focused on one citable idea when possible.
- Preserve original order and stable segment IDs such as `S001`, `S002`, `S003`.
- Skip obvious non-citable connective sentences unless the user asks to cite every sentence.
- For very long text, process in batches but keep a single final mapping table.
Default segmentation rules: use blank lines as paragraph boundaries; if a paragraph is longer than about 700 characters or contains multiple claims, split into sentences; merge very short fragments into neighboring text unless they contain a distinct claim; keep section headings as labels, not as citable segments. If the input has more than about 10 segments, prefer batch mode.
## 2. Parse each segment
For each citable segment:
- Extract the core claim in one sentence.
- Identify claim type: `mechanism`, `association`, `method`, `clinical`, `epidemiology`, `background`, `definition`, or `review-context`.
- Identify entities, intervention/exposure, outcome, population/model, directionality, and boundary.
- Convert the claim into 2-4 English search queries: one precise query with all key terms; one synonym query; one broader background query; one methods or model query if relevant.
If the claim is too broad, split it into citable subclaims rather than searching the whole sentence. For deeper help turning a claim into queries and support grades, open `references/search-strategy.md`.
## 3. Search candidate papers
Prefer `scripts/nature_citation.py` when internet access is available. The full flag list, polite-pool/rate-limit options, and the long-article batch strategy are in `references/script-usage.md`. A minimal run:
```bash
python scripts/nature_citation.py \
--text "PASTE MANUSCRIPT TEXT HERE" \
--scope cns \
--outdir /tmp/nature-citation \
--format ris \
--with-artifacts
```
When the topic is biomedical or PubMed-indexed, also search PubMed with journal filters and compare results against Crossref. Use NCBI E-utilities rate limits and include `tool`/`email` parameters if running repeated searches.
## 4. Evaluate whether each paper supports the segment
Use a conservative support scale:
- `strong support`: the paper directly tests the same relationship/mechanism/method and the result supports the segment.
- `partial support`: the paper supports part of the segment, a related model, or a narrower condition.
- `background support`: the paper supports field context, not the specific claim.
- `contradictory/limiting`: the paper conflicts with or narrows the claim.
- `metadata-only candidate`: title/metadata suggest relevance, but abstract/full text has not been checked.
Never cite a `metadata-only candidate` as support without checking the abstract or publisher page. If a paper is a review, label it as review/context and avoid using it as primary evidence for an experimental claim when primary articles are available.
## 5. Export reference-manager file
Default behavior: write one reference-manager file; support publication time filters with `--from-year` and `--to-year`; for long or ambiguous texts, use `--with-artifacts` so the HTML browser is available.
Default file is `references.ris`. Optionally write `references.enw` for EndNote tagged import or `references.rdf` for Zotero RDF. If the user asks to choose the download format, treat `RIS`, `ENW`, and `Zotero RDF` as the supported options and return only one export file unless they explicitly ask for multiple. Do not invent missing fields. If DOI, pages, volume, or issue are missing, leave them absent.
Before writing any reference-manager file, run the author-integrity gate.
- Use the complete ordered author list from structured Crossref, PubMed, or publisher metadata. Never derive `AU` fields from a display citation, an `et al.` string, or a surname-only list.
- Write one `AU` line per person in `Family, Given` form and retain initials, suffixes, particles, and ordering supplied by the source.
- Encode a consortium or institutional author as a corporate author with a trailing comma so EndNote does not invert it.
- If any personal author lacks a family name or given name/initials, stop the export. Refetch by DOI or PMID and compare against the publisher record. Do not label the file EndNote-ready while the author list is incomplete.
- Permit `--allow-incomplete-authors` only as an explicit emergency override, and keep warnings in the exported record.
See `references/ris-endnote.md` for format details and the EndNote import option.
## 6. Optional review artifacts
Generate review artifacts (HTML/TSV/JSON/report) for long or ambiguous runs — they are the primary way the user browses, filters, and selects candidates. Use `--with-artifacts` when the text is long, the query is broad, or the user needs manual curation. Report the HTML visualization path prominently when artifacts are enabled, and generate TSV/JSON/report alongside the HTML.
## 7. Report results
Unless the user asks for a different format, return:
```text
交互式引用浏览器
- [absolute path to citation_visualization.html] ← 在浏览器中打开此文件，可筛选/选择/下载引用
检索范围
- [Nature Portfolio / Science family / Cell Press / flagship only, plus date limits]
分段引用对应关系
S001: [source segment]
- [Author, year, title, journal, DOI]
- 支撑等级: [strong/partial/background/limiting/metadata-only]
- 插入建议: [e.g. after sentence / after clause]
导出文件
- [absolute path to references.ris / references.enw / references.rdf]
风险和缺口
- [missing full-text check, contradictory evidence, no direct CNS literature, etc.]
```
Put the HTML browser path FIRST, above everything else, so the user can immediately open and browse candidates. If no suitable CNS/Nature-series paper exists, say so plainly and suggest the best nearby options from non-CNS literature only if the user wants broader coverage. If the text is long, mention the batch strategy used, especially when you limited the run with `--batch-size` or `--max-segments`.
tests/test_author_exports.py›
from __future__ import annotations
import argparse
import importlib.util
import sys
import tempfile
import unittest
from unittest import mock
from pathlib import Path
import xml.etree.ElementTree as ET
SCRIPT = Path(__file__).parents[1] / "scripts" / "nature_citation.py"
SPEC = importlib.util.spec_from_file_location("nature_citation_under_test", SCRIPT)
assert SPEC and SPEC.loader
MODULE = importlib.util.module_from_spec(SPEC)
sys.modules[SPEC.name] = MODULE
SPEC.loader.exec_module(MODULE)
def candidate(**overrides):
values = {
"title": "A complete reference",
"journal": "Nature Medicine",
"family": "Nature Portfolio",
"year": "2024",
"y1": "2024/01/01",
"doi": "10.1000/example",
"url": "",
"volume": "1",
"issue": "2",
"start_page": "10",
"end_page": "18",
"issn": "0000-0000",
"authors": ["Smith, Alfred", "NMSS Validation Group,"],
"abstract": "",
"type": "journal-article",
"score": 0.0,
"source_query": "test",
}
values.update(overrides)
return MODULE.Candidate(**values)
class AuthorExportTests(unittest.TestCase):
def test_ris_is_the_default(self):
self.assertEqual(MODULE.DEFAULT_EXPORT_FORMAT, "ris")
self.assertEqual(MODULE.infer_export_format(None), "ris")
self.assertEqual(MODULE.export_filename("ris"), "references.ris")
def test_crossref_preserves_given_names_suffixes_and_corporate_authors(self):
item = {
"title": ["Author metadata test"],
"container-title": ["Nature Medicine"],
"published": {"date-parts": [[2024, 2, 3]]},
"DOI": "10.1000/authors",
"author": [
{"family": "Smith", "given": "Alfred", "suffix": "Jr."},
{"name": "NMSS Validation Group"},
],
}
result = MODULE.candidate_from_crossref(item, "doi:10.1000/authors")
self.assertIsNotNone(result)
self.assertEqual(result.authors, ["Smith, Alfred, Jr.", "NMSS Validation Group,"])
self.assertEqual(result.author_warnings, [])
def test_incomplete_personal_author_is_blocked_by_default(self):
incomplete = candidate(
authors=["Chaudhuri"],
author_warnings=["Author 1 lacks a given name or initials in structured metadata."],
)
with self.assertRaisesRegex(ValueError, "Author metadata integrity check failed"):
MODULE.validate_author_integrity([incomplete])
MODULE.validate_author_integrity([incomplete], allow_incomplete=True)
unflagged_surname = candidate(authors=["Chaudhuri"], author_warnings=[])
with self.assertRaisesRegex(ValueError, "complete personal name"):
MODULE.validate_author_integrity([unflagged_surname])
def test_ris_has_one_author_per_line_and_traceable_identifiers(self):
record = MODULE.build_ris_record(candidate(pmid="12345678"))
self.assertIn("AU - Smith, Alfred", record)
self.assertIn("AU - NMSS Validation Group,", record)
self.assertEqual(record.count("AU - "), 2)
self.assertIn("DO - 10.1000/example", record)
self.assertIn("AN - PMID:12345678", record)
def test_write_ris_requires_explicit_override_for_incomplete_authors(self):
incomplete = candidate(authors=["Smith"], author_warnings=["surname only"])
with tempfile.TemporaryDirectory() as directory:
output = Path(directory) / "references.ris"
with self.assertRaises(ValueError):
MODULE.write_ris([incomplete], output)
MODULE.write_ris([incomplete], output, allow_incomplete_authors=True)
self.assertIn("Author metadata warning: surname only", output.read_text(encoding="utf-8"))
def test_pubmed_parser_retains_personal_and_collective_authors(self):
xml = """
<PubmedArticle>
<MedlineCitation>
<PMID>21264941</PMID>
<Article>
<ArticleTitle>Validation of a nonmotor symptoms questionnaire.</ArticleTitle>
<Abstract><AbstractText>Abstract text.</AbstractText></Abstract>
<Journal>
<ISSN>0000-0000</ISSN>
<JournalIssue>
<Volume>26</Volume><Issue>10</Issue>
<PubDate><Year>2010</Year><Month>Oct</Month><Day>5</Day></PubDate>
</JournalIssue>
<Title>Nature Medicine</Title>
</Journal>
<Pagination><MedlinePgn>123-130</MedlinePgn></Pagination>
<AuthorList>
<Author><LastName>Chaudhuri</LastName><ForeName>K Ray</ForeName></Author>
<Author><LastName>Schapira</LastName><ForeName>Anthony H V</ForeName><Suffix>Jr.</Suffix></Author>
<Author><CollectiveName>NMSS Validation Group</CollectiveName></Author>
</AuthorList>
<PublicationTypeList><PublicationType>Journal Article</PublicationType></PublicationTypeList>
</Article>
</MedlineCitation>
<PubmedData><ArticleIdList><ArticleId IdType="doi">10.1000/pubmed</ArticleId></ArticleIdList></PubmedData>
</PubmedArticle>
"""
result = MODULE.candidate_from_pubmed(ET.fromstring(xml))
self.assertIsNotNone(result)
self.assertEqual(
result.authors,
["Chaudhuri, K Ray", "Schapira, Anthony H V, Jr.", "NMSS Validation Group,"],
)
self.assertEqual(result.author_warnings, [])
self.assertEqual(result.pmid, "21264941")
self.assertEqual(result.doi, "10.1000/pubmed")
self.assertEqual(result.y1, "2010/10/05")
def test_pubmed_candidate_replaces_incomplete_crossref_duplicate(self):
crossref = candidate(authors=["Smith"], author_warnings=["surname only"])
pubmed = candidate(
authors=["Smith, Alfred"],
author_warnings=[],
pmid="12345678",
source_catalog="PubMed",
)
result = MODULE.dedupe([crossref, pubmed])
self.assertEqual(len(result), 1)
self.assertEqual(result[0].source_catalog, "PubMed")
self.assertEqual(result[0].authors, ["Smith, Alfred"])
def test_read_pmids_normalizes_urls_and_rejects_invalid_values(self):
args = argparse.Namespace(
pmid=["PMID: 123", "https://pubmed.ncbi.nlm.nih.gov/456/", "123"],
pmid_file=None,
)
self.assertEqual(MODULE.read_pmids(args), ["123", "456"])
args.pmid = ["not-a-pmid"]
with self.assertRaisesRegex(ValueError, "Invalid PMID"):
MODULE.read_pmids(args)
def test_failed_explicit_pmid_does_not_write_empty_success_export(self):
with tempfile.TemporaryDirectory() as directory:
output = Path(directory) / "references.ris"
with mock.patch.object(
MODULE,
"fetch_pubmed_candidates",
return_value=([], [{"pmid": "123", "error": "network unavailable"}]),
):
code = MODULE.main(["--pmid", "123", "--output-file", str(output), "--sleep", "0"])
self.assertEqual(code, 4)
self.assertFalse(output.exists())
if __name__ == "__main__":
unittest.main()