视觉画布设计
canvas-design
基于设计原则创作静态视觉作品,用于海报与品牌图形。
academic-docs【文档与办公】Operate GPT Academic paper, PDF, Arxiv, LaTeX, Word, Markdown, and batch document workflows.
独立 GitHub 快照 · 自动分类 · 以原作者文档为准
Use this sub-skill when a GPT Academic task consumes scholarly or office documents: PDF papers, Arxiv IDs, DOI-like paper requests, LaTeX projects, Word files, Markdown files, Google Scholar pages, batch document folders, or paper-reading prompts.
Read this sub-skill for: “PDF论文翻译”, “ChatPDF”, “PDF QA”, “Arxiv论文翻译”, “速读论文”, “论文摘要”, “LaTeX润色/纠错/翻译”, “Word总结”, “DOC2X”, “GROBID”, “NOUGAT”, “Google Scholar”, “批量文件询问”, “paper reading”, “translate this paper”, or “summarize uploaded documents”.
python scripts/check_doc_backends.py --repo-root <checkout>
python sub-skills/academic-docs/scripts/check_document_input.py <document-or-folder>
references/configuration.md or ../conversation/SKILL.md.| User goal | GPT Academic workflow | Read next |
|---|---|---|
| translate a PDF paper | PDF论文翻译; parser order DOC2X → GROBID → traditional | references/workflows.md, references/document-backends.md |
| ask questions about a PDF | 理解PDF文档内容 (模仿ChatPDF) | references/workflows.md |
| translate Arxiv paper | Arxiv论文翻译 or LaTeX-backed Arxiv plugins | references/workflows.md |
| polish/proofread/translate LaTeX | LaTeX project polish/proofread/translate plugins | references/workflows.md, references/document-backends.md |
| summarize Word/Markdown/folder | Word summary, Markdown translation, batch file query | references/workflows.md |
| formula-heavy PDF OCR | 精准翻译PDF文档(NOUGAT) | references/document-backends.md |
| academic search/paper discovery | Google Scholar assistant or academic conversation | references/workflows.md and ../conversation/SKILL.md |
../conversation/SKILL.md unless a document workflow is the primary input.../programming-code/SKILL.md.../agent-tooling/SKILL.md.../multimodal-media/SKILL.md.Native candidates include DOC2X parsing, LaTeX auto-correct, Markdown conversion, safe pickle for LaTeX state, and plugin harness examples. Many require network, credentials, LaTeX binaries, or LLM APIs; do not treat skipped optional parser tests as passing.