# Read and translate academic papers bilingually

Reading academic papers in a foreign language is slow and imprecise. This skill converts PDFs, DOIs, arXiv papers, and HTML articles into complete bilingual Markdown readers with figures, tables, and source-grounded citations.

## Install

```bash
npx skillstore add community contribution, refactored into static/dynamic layers/yuan1z0825-nature-reader
```

## Metadata

- - Slug: yuan1z0825-nature-reader
- - Version: 2.0.0
- - Author: Community contribution, refactored into static/dynamic layers
- - GitHub username: Yuan1z0825
- - License: MIT
- - Repository: https://github.com/Yuan1z0825/nature-skills/tree/main/skills/nature-reader
- - Ref: main
- - Supported tools: Claude, Codex, Claude Code
- - Risk level: safe
- - Quality score: 83
- - Quality tier: gold
- - Public page: https://skillstore.pages.dev/skills/yuan1z0825-nature-reader
- - Manifest: https://skillstore.pages.dev/api/skills/yuan1z0825-nature-reader/manifest

## Capabilities

- Convert academic papers into paragraph-level bilingual Markdown readers with original text and Chinese translation side by side
- Extract figures and tables from PDFs and place them near their first substantive mention in the translation
- Maintain stable source block IDs \(S001, F001, T001\) for every paragraph, figure, and table for traceability
- Support multiple input formats: PDF text, scanned PDFs, publisher HTML, DOI/arXiv links, and pasted text
- Generate output files including paper.md, source\_map.json, translation\_notes.md, and an assets folder
- Preserve technical terms, equations, citations, and numeric values during translation

## Use Cases

- Graduate student reading a Chinese-language paper: A graduate student needs to understand a Chinese academic paper and wants bilingual output to verify translation accuracy against the original.
- Researcher translating a paper for a colleague: A researcher wants to share a translated version of an important paper with a colleague who reads Chinese, maintaining figure positions and citations.
- Literature review assistant: A PhD candidate conducting literature review needs to process multiple papers efficiently while maintaining traceability back to source blocks for follow-up questions.

## Prompt Templates

### Basic PDF translation

```
Translate this paper into a bilingual reader: [paste PDF or attach file]. Keep figures and tables near their discussion sections.
```

### DOI or arXiv resolution

```
Build a bilingual reader for this paper: https://doi.org/10.1038/s41586-2024-12345 or arXiv:2401.12345. Resolve the link first, then translate.
```

### Pasted text processing

```
Here is pasted text from a paper [paste content]. Create a bilingual Markdown reader with stable block IDs for each paragraph.
```

### Follow-up question with source grounding

```
Looking at the paper.md you created, what does Section 3 say about the model's training data? Cite the block ID and page number.
```

## Limitations

- Cannot reproduce large copyrighted publisher text directly in chat responses — must be saved to local artifact files
- OCR quality for scanned PDFs depends on the source document clarity; uncertain blocks are marked rather than guessed
- Requires the full document to produce a complete reader; will not summarize unless explicitly requested
- Translation quality depends on the underlying language model capabilities

## Best Practices

- Always process the full document when a complete paper is provided — do not stop at the abstract or introduction
- State the detected source\_format \(pdf-text, scanned-pdf, html, doi-arxiv, or pasted-text\) before processing so users can correct early
- Mark uncertain blocks in translation\_notes.md rather than guessing when OCR or layout extraction is weak
- Keep the Terminology Ledger consistent across the whole document for recurring technical terms

## Anti Patterns

- Do not degrade a translation request into a summary-only output unless the user explicitly asks for a summary
- Do not convert dense prose into bullet-point keywords — preserve paragraph form
- Do not silently skip Methods, limitations, data availability, or competing interests sections
- Do not reproduce large copyrighted text directly in chat — save to local artifact files

## Security Audit

- - Safe to publish: true
- - Audited at: 2026-06-24T06:06:24.158\+00:00
- - Summary: The nature-reader skill is a documentation-only skill consisting entirely of Markdown instruction files and a YAML manifest. There is no executable code, no network requests, no scripts, and no environment access. All 170 static findings are false positives: the 'Ruby/shell backtick execution' patterns are Markdown inline code formatting, 'path traversal' patterns are relative cross-reference paths in documentation, 'weak cryptographic' findings are block ID naming conventions, and 'system reconnaissance' findings are file structure descriptions. No security risk was identified.

## Stats

- - Views: 0
- - Downloads: 20
- - Favorites: 1
- - Popularity score: 0