Full-Process Academic Writing and Document Production

Paper production as an engineering workflow

Academic writing often breaks down between research, figures, and formatting: notes are scattered, Matplotlib plots do not match publication styling, and the final manuscript still has to be hand-assembled into LaTeX or Word. This skill turns research → figures → manuscript → output into an executable pipeline. It starts by clarifying output format, task type, figure scope, target venue, language, and acceptance criteria before executing.

From search to LaTeX/PDF or .docx

The core path covers:
- Literature search: field-specific WebSearch queries, WebFetch extraction of methods, data, and citations, written to research_notes.md.
- Figure generation: reads plot_types.md and styling_guide.md, then generates at least nine applicable figures/ using plot_template.py.
- IMRAD writing: reads structure and citation references, then drafts full prose sections such as Abstract, Methods, Results, and Discussion.
- Output: the LaTeX/PDF track writes manuscript.tex, embeds PNGs, and compiles with xelatex; the Word track uses new_doc.js to produce a Chinese-academic .docx with three-line tables, OMML formulas, chapter numbering, and GB/T 7714 references.

Boundaries: if data or figures already exist, search and plotting can be reduced; if the task is editing .docx or converting Markdown to .docx, unnecessary steps should be skipped. If LaTeX is unavailable, the skill delivers .tex rather than installing dependencies. Word is better for Chinese submissions; LaTeX/PDF is better for international venues.