Yuan1z0825/nature-skills

This system is a modular library of reusable research and academic writing skills designed to automate and enforce quality in scholarly workflows. It provides specialized tools for academic search, citation management, figure generation, and patent drafting, alongside structured capabilities for manuscript preparation, peer-review rebuttals, and slide reconstruction. The infrastructure ensures reliability through rigorous automated validation, metadata consistency checks, and safety gates for external API interactions and content generation.

The \nature-academic-search\ skill now includes a standalone, no-MCP fallback implementation in \scripts/academic\_search.py\ that queries the OpenAlex API for academic papers by topic, author, ORCID, or institution, supporting name disambiguation and result re-ranking. Additionally, \scripts/format-converter.py\ and \scripts/converters.py\ provide a self-contained workflow to download citations from PubMed, CrossRef, and arXiv and convert them into RIS, BibTeX, or ENW formats, while \scripts/preflight.py\ allows users to verify connectivity to these external APIs before running searches or downloads.

skills/nature-academic-search/scripts · high confidence

Added a set of reference documents in the \skills/nature-academic-search/references\ directory that define the operational logic for the new academic search skill. These include specifications for a citation parser (extracting DOIs, PMIDs, and arXiv IDs from .docx, .tex, and .txt files), a deduplication engine (using DOI and Jaccard similarity for title/author matching), and source tier classifications (T1-T3) for routing searches across PubMed, CrossRef, arXiv, and others. The entry also introduces six workflow definitions: multi-source search, citation verification, MeSH-based query construction, citation file management (format conversion), general reference management (related papers, BibTeX generation), and a strict other-citation impact audit workflow that distinguishes independent citations from self/team citations and generates article-level metric tables.

skills/nature-academic-search/references · high confidence

A new Python script, build\_proposal\_docx.py, has been added to the nature-proposal-writer skills to convert research proposal content from Markdown into professionally formatted .docx files. This tool applies specific academic styling standards, including Times New Roman and SimSun fonts, 1.5 line spacing, 1-inch margins, and specific heading sizes, ensuring that generated proposals meet formal submission requirements.

skills/nature-proposal-writer/scripts · high confidence

Added four new LaTeX templates in the skills/nature-writing/templates/submission directory to support the initial manuscript submission workflow. These templates include a title page (title-page.tex) for author details, keywords, and mandatory declarations; an initial cover letter (initial-cover-letter.tex) with placeholders for editor and journal information; a highlights and editorial summary template (highlights-and-summary.tex) for principal findings; and a declarations and reviewers template (declarations-and-reviewers.tex) for suggested/opposed reviewers and compliance statements. All templates use AUTHOR\_INPUT\_NEEDED placeholders to guide authors in providing required content.

skills/nature-writing/templates · high confidence

The new \nature-academic-search\ skill enables multi-source scholarly search, citation verification, and strict external-citation auditing. It uses a static/dynamic router architecture defined in \manifest.yaml\ and \SKILL.md\ to dynamically load specific workflow fragments (such as multi-source search, MeSH strategy building, or citation-file management) based on user intent, rather than executing a monolithic process. The skill integrates with an MCP server for API-driven retrieval from sources like PubMed, CrossRef, and arXiv, while providing a no-MCP fallback via \scripts/academic\_search.py\ for discovery when the server is unavailable. An installer (\install.sh\) automates the setup of the MCP server, Python dependencies, and Claude Code configuration, and the skill supports bilingual documentation in English and Chinese.

skills/nature-academic-search · high confidence

Adds the \nature-downloader\ skill, which routes literature downloads through lawful open-access, publisher-API, CNKI institutional, and browser-based institutional providers. The skill requires explicit user confirmation for downloading Supporting Information (SI) and enforces strict boundaries: it does not bypass paywalls, DRM, or CAPTCHAs, and it never reads or exports browser cookies, passwords, or session files. It supports configuration of library resource URLs and publisher API keys (Elsevier, Springer Nature, IEEE) and provides an auditable manifest for every download attempt.

skills/nature-downloader · high confidence

The \nature-image2ppt\ skill is now available to reconstruct existing slide images, screenshots, scanned PDFs, or image-only PPTX files into object-level editable PowerPoint decks. It restores existing visual pages rather than authoring new presentations from notes, rebuilding text boxes, shapes, connectors, and formulas as native objects while preserving speaker notes. The skill uses a deterministic CLI workflow (\python cli/image2ppt/cli.py\) with a v2 page manifest system, atomic finalization to prevent partial outputs, and a provider-neutral image backend that supports both built-in generation and third-party OpenAI-compatible APIs. It requires Python 3.10+, uses Baidu AI Studio for online OCR (with a local fallback), and includes comprehensive QA checks to ensure visual fidelity.

skills/nature-image2ppt · high confidence

The \nature-paper-card\ skill is now available to perform deep, structured reading of single scientific papers, generating a source-grounded 16-section Paper Card. It focuses on reconstructing the paper's argument, mapping evidence to claims, identifying conclusion boundaries, and proposing testable research ideas, rather than producing summaries or translations. The skill supports three locator modes—page-grounded, structure-grounded, and source-limited—to handle varying levels of source reliability, and includes bundled scripts for preparing source bundles and auditing the output for structural integrity and provenance accuracy.

skills/nature-paper-card · high confidence

The \nature-response\ skill is now available to help authors draft, audit, and revise peer-review rebuttal packages. It supports both Major and Minor revisions by parsing editor decision letters and reviewer reports to generate mutually isolated, point-by-point responses for each reviewer, ensuring no cross-reviewer information leaks. The skill includes a structured internal tracker to manage response strategies, evidence requirements, and submission readiness, while strictly preventing the fabrication of data or line numbers. It also provides automated consistency checks for LaTeX revision packages to ensure that response letters, marked manuscripts, and clean manuscripts remain synchronized, and includes bilingual support for Chinese authors to align on missing facts and high-risk issues.

skills/nature-response · high confidence

A new \nature-shared\ support package has been added to centralize definitions and reference materials used by multiple \nature-\*\ skills (such as \nature-writing\ and \nature-polishing\). This package includes shared core guidelines for manuscript consistency, discussion argumentation, main-text discipline, and Nature-style abstract/introduction structures, along with journal-specific format references for flagship Nature and Nature Machine Intelligence. It also provides a \check\_consistency.py\ script for mechanical terminology and numeric audits. The package is designed to be installed as a dependency and loaded by other skills, rather than invoked as a standalone user workflow.

skills/nature-shared · high confidence

Adds a new MCP server that provides a unified search tool across CrossRef, PubMed, and arXiv, along with dedicated tools for Scopus and ScienceDirect. The unified search defaults to the free sources (CrossRef, PubMed, arXiv) to avoid consuming Elsevier API quotas; Scopus and ScienceDirect are available only when explicitly requested via the \sources\ parameter or by calling their specific tools, requiring a local \pybliometrics\ configuration file. The server also includes tools for fetching paper details, formatted citations, and MeSH lookups, and hardens the arXiv source by disabling external XML entity processing.

skills/nature-academic-search/mcp-server · high confidence

A new skill has been added to convert research papers, technical reports, source code, and inventor notes into evidence-grounded Chinese invention patent drafts and technical disclosure documents (技术交底书). The skill provides a structured workflow that includes scanning project materials, mining patent points, performing prior-art searches (prioritizing CNIPA databases), and generating formal deliverables such as claims, specifications, abstracts, and Mermaid-based system/process diagrams in both Markdown and DOCX formats. It enforces strict evidence discipline by mapping claim features to source IDs and marking unsupported features as gaps, while also supporting iterative refinement of disclosure drafts with timestamped versioning.

skills/nature-paper-to-patent · high confidence

This change introduces a suite of new scripts to automate skill updates and enforce documentation and metadata quality. The \autoupdate-skills.sh\ script enables automatic, throttled synchronization of installed skills from the repository to destinations like \\~/.claude/skills\ or \\~/.codex/skills\, handling state isolation and lock management. The \update-codex-skills.sh\ script provides a manual or scripted way to install and update skills into Codex with transactional safety and pruning of stale entries. Documentation integrity is enforced by \validate-readmes.py\ (checking badge/index consistency), \validate-readme-mirror.py\ (ensuring Chinese/English README alignment), and \update-reference-tocs.py\ (auto-generating navigation lists in long reference files). Metadata and workflow integrity are validated by \validate-skill-metadata.py\ (checking frontmatter, manifest routes, and Codex interface contracts), \validate-skill-index.py\ (ensuring the README index matches the skills directory), and \validate-workflows.py\ (verifying CI path filters and pinned action SHAs). Additionally, \generate-star-history.py\ creates static SVG star-history charts from GitHub data.

scripts · high confidence

This change introduces a suite of new Python scripts to enforce quality and consistency across the repository and individual skills. A new \scripts/validate-repository.py\ tool validates the overall repository structure, checking for missing skill files, malformed JSON/TOML configs, broken README links, and incorrect skill count badges. Specific skills gain dedicated validation and auditing tools: \nature-citation\ now includes a script to segment text and export citations in RIS format; \nature-figure\ adds scripts to audit PDF text font sizes, detect geometric collisions in rendered figures, and perform static preflight checks on Python and R source code; \nature-paper-to-patent\ introduces scripts to audit Chinese patent claims for structural integrity and validate patent drafts for traceability and completeness; \nature-paper2ppt\ adds a lightweight PPTX quality auditor to check for layout defects and AI-template phrases; and \nature-reader\ includes a script to validate equation rendering and traceability. These additions provide automated checks to catch errors before merging.

python · high confidence

The project has transitioned from a single \nature-polishing\ skill to a comprehensive library of 19 reusable research skills (\nature-\*\), complete with bilingual documentation (Chinese and English READMEs) and a dedicated website. The license has been upgraded from MIT to Apache 2.0. To ensure data integrity, the repository now includes automated CI workflows that validate skill metadata, verify that READMEs are mirrored across languages, and check that the skill index matches the actual directory structure.

(repo-wide) · high confidence

The \nature-figure\ skill has been restructured around a new router and static/dynamic architecture that enforces stricter quality and safety standards for scientific figure generation. Users now benefit from a backend-selection gate that explicitly routes tasks to either Python (matplotlib/seaborn) or R (ggplot2/patchwork) and enforces exclusive usage for all drawing and QA steps. The skill adds mandatory render-time panel-alignment and collision-audit gates: multi-panel figures are automatically checked for geometric alignment (with a 1.5 pt tolerance for shared edges and gutters) and text/shape collisions in the final PDF, blocking export if critical issues are found. It also introduces a dedicated non-plotting route for AI-generated graphical abstracts and mechanism schematics via OpenRouter (GPT Image 2), complete with policy gates and provenance checks, while retaining third-party \figures4papers\ reference materials with clear copyright boundaries.

skills/nature-figure · high confidence

Added automated tests for skill update safety, writing guidance contracts, and figure generation audits; Added test fixtures and contract verification for reviewer behavior; Added tests for author metadata handling and RIS export in the Nature Citation skill; Added tests for star history generation script.

This change introduces explicit dependency files for several new skills and the CI environment. A pinned \requirements-ci.txt\ is added to \.github/\ to ensure reproducible Python 3.11 test environments in GitHub Actions. New \requirements.txt\ files are added for the \nature-academic-search\ (MCP server), \nature-downloader\, \nature-figure\, \nature-image2ppt\, and \nature-paper-to-patent\ skills, specifying their Python dependencies (e.g., \mcp\, \pybliometrics\, \pdfplumber\, \PyMuPDF\, \openai\, \latex2mathml\). Additionally, \package.json\ and \package-lock.json\ are added for the \nature-paper-to-patent\ disclosure script to manage Node.js dev dependencies (\@mermaid-js/mermaid-cli\, \puppeteer\), and a \requirements-cnipa.txt\ is added for the CNIPA search script (requiring \playwright\).

(dependencies) · high confidence

Written by watchdog.canine.dev from the codebase's own history, inside the signed delivery this page is composed from.

• First survey — no prior run to compare against. CAI 57.

• Code Health 48
• Architecture 91
• Maturity 73
• Readiness 53
• Security 87
• Accessibility 74

• 300 commits — 248 feature/other, 52 fixes

• (root) — 100 commits
• (repo) — 85 commits
• .github/workflows — 15 commits
• skills/nature-figure — 15 commits
• skills/nature-downloader — 9 commits
• skills/nature-response — 8 commits
• skills/nature-image2ppt — 7 commits
• skills/nature-academic-search — 6 commits
• skills/nature-writing — 5 commits
• scripts/autoupdate-skills.sh — 4 commits
• scripts/tests — 4 commits
• skills/nature-shared — 4 commits
• assets/star-history.svg — 3 commits
• skills/nature-paper-to-patent — 3 commits
• skills/nature-polishing — 3 commits
• scripts/generate-star-history.py — 2 commits
• scripts/update-codex-skills.sh — 2 commits
• scripts/validate-repository.py — 2 commits
• scripts/validate-skill-metadata.py — 2 commits
• skills/nature-citation — 2 commits

• fix: Merge pull request #100 from subhansh-dev/fix/parse-scalar-numeric-bool
• fix: Merge pull request #123 from Travisma2233/fix/star-history-cache-busting
• fix: Merge pull request #128 from Travisma2233/docs/fix-directory-layout-level
• fix: Merge pull request #132 from Travisma2233/fix/submission-reviewer-table
• fix: Merge pull request #133 from Travisma2233/docs/fix-self-operated-section
• fix: Merge pull request #137 from Travisma2233/fix/interactive-installation-section
• fix: Merge pull request #139 from Travisma2233/fix/mobile-category-active-text
• fix: Merge pull request #142 from Travisma2233/docs/fix-readme-contents-hierarchy
• fix: Merge pull request #159 from zpf2234/fix/autoupdate-throttle-validation
• fix: Merge pull request #160 from zpf2234/fix/update-codex-dest-equals
• fix: Merge pull request #166 from Yuan1z0825/fix/repository-skill-count-validation
• fix: Merge pull request #168 from zpf2234/fix/validate-manifest-route-paths
• fix: Merge pull request #212 from zpf2234/fix/windows-mermaid-command
• fix: Merge pull request #216 from zpf2234/fix/patent-omml-superscript-element
• fix: Merge pull request #217 from zpf2234/fix/atomic-image2ppt-config
• fix: Merge pull request #222 from anupamme/fix-repo-nature-skills-gitlab-bandit-b313-b314-b315-b316-b318-b319-b320-b405-b406-b407-b408-1e4267ad
• fix: Merge pull request #223 from anupamme/fix-repo-nature-skills-v-003-page-dir-path-traversal
• fix: Merge pull request #224 from anupamme/fix-repo-nature-skills-pubmed-xxe-defusedxml
• fix: fix(academic-search): use ESummary JSON for MeSH lookup (#214)
• fix: fix: V-003 security vulnerability
• …and 280 more

• 0 containers · 1 bounded contexts · 0 dependency edges (baseline)

Written by watchdog.canine.dev from the codebase's own history, inside the signed delivery this page is composed from.

Yuan1z0825/nature-skills was measured the same way every project in this corpus was: the same rubric, at a pinned commit, with the result published in full. Point a surveyor at a repository you know and see whether you agree with it.