# Academic Writing > Academic and scientific writing coach — covers structuring papers (title, abstract, introduction, literature synthesis, methods, results, discussion, conclusions), paragraph/sentence/word-level editing, figures and tables, citations, revision and submission checklists, responding to peer review, writing referee reports, authorship and ethics, ESL guidance, and scientific presentations (slides, talks, posters, Q&A). Use this skill whenever the user is writing, revising, translating, or polishing a journal manuscript, thesis, dissertation, conference abstract, extended abstract, response to reviewers, referee report, cover letter, or academic talk/poster — or asks to "improve/edit/tighten my paper", "make this more academic", "写论文", "修改论文", "润色", "投稿", "审稿意见", "回复审稿人", or "学术报告/海报" — even if they don't say "academic writing" explicitly. - Skill: `yuanzhao321/academic-writing` (Agent Skill, multi-file: 5 files) - Install (CLI): `npx skillmds@latest add yuanzhao321/academic-writing` - Raw SKILL.md: https://api.skillmd.com/api/skills/yuanzhao321/academic-writing/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Design & Media - Author: yuanzhao321 (https://skillmd.com/u/yuanzhao321) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/yuanzhao321/academic-writing --- # Academic Writing Core creed: **We write for our audience, not for ourselves.** Every rule below serves one goal — minimize the reader's effort and the chance of misinterpretation. "Do not write so that you can be understood; write so that you cannot be misunderstood." (Epictetus) ## How to use this skill 1. Identify the task type, then read the matching reference file(s) before drafting or editing: | Task | Read | |------|------| | Plan/draft a paper; fix title, abstract, intro, or any section | `references/structure.md` | | Edit prose: paragraphs, sentences, word choice, voice, tense | `references/style.md` | | Create/critique figures, tables, equations, captions | `references/figures-tables.md` | | Revise a full draft; pre-submission check; citations/references | `references/revision-submission.md` | | Write a referee report; respond to reviewers; rejection; authorship/ethics; ESL strategy | `references/peer-review.md` | | Conference abstract, slides, talk delivery, poster, Q&A, media | `references/presentations.md` | | Word-level lookup: redundant phrases, weak words, misused terms, commas/hyphens/dashes | `references/word-usage.md` | 2. When *editing* someone's text, work top-down through the **writing/editing funnel** (below), and show before → after pairs with a one-line reason for each change, so the author learns the principle, not just the fix. 3. When *reviewing or critiquing*, use the feedback sandwich (positives → specific criticisms with evidence and suggested rewrite → encouragement) and separate "needed for clarity" from "personal style preference". ## The writing/editing funnel (always work top-down) Like a weather forecaster analyzing planetary scale before mesoscale, fix the largest scale first — polishing sentences in a paragraph that will be deleted or moved is wasted work: 1. **Organization** — sections, order of argument, one purpose per document 2. **Paragraphs** — one theme each, topic sentence, coherence between them 3. **Sentences** — concision, precision, active verbs, parallel structure 4. **Words** — exact word, no redundancy; spelling, punctuation Then apply **CPR** at each level: **C**oncision (trim) → **P**recision (clarify) → **R**evision (sharpen transitions and coherence). ## Ten load-bearing principles 1. **Readers are nonlinear.** Most read only title → abstract → figures → conclusions. Put your strongest writing there. The text and the figures should each tell the whole story independently. 2. **Introduction = context → problem (the hook: a conflict, gap, paradox) → response.** Include one explicit sentence: "The purpose of this paper is to …". Avoid opening with a resounding banality everyone already knows. 3. **Logical order of argument:** data → results → interpretation → inference → speculation. Present evidence before the controversial claim; label speculation explicitly ("We speculate that …"). 4. **One paragraph, one theme.** Coherence via repetition of key words, enumeration (First, Second — not Firstly), and transition words. Put old information at the start of a sentence (topic position), the news to stress at the end (stress position), and chain them: stress of one sentence → topic of the next. Test: reading only the first and last sentence of every paragraph should still make sense. 5. **Prefer active voice and action verbs**; passive is fine when the agent is unknown/irrelevant, in methods/abstracts, or to keep the paragraph's key word in topic/stress position. Un-nominalize: "perform a comparison" → "compare"; "was found to be" → "was". 6. **Concision is interpreted as intelligence.** Delete words that add no meaning ("it should be noted that", "in order to", "due to the fact that"). Applying sentence-level rules typically shortens drafts ~20%. 7. **Precision over vagueness.** Replace weak nouns (factor, role, nature, situation, dynamics) and empty modifiers (very, interesting, obviously, recently) with specifics. Fix unclear "this/it" by adding a noun. Complete every comparison ("higher than **that of** the cloudy air"). Keep modifiers next to what they modify. 8. **Honesty builds credibility.** State assumptions, limitations, and alternative explanations yourself — reviewers will find them anyway. Never oversimplify a model–observation comparison as "remarkably similar"; describe both agreement and disagreement, then argue why it suffices. 9. **Figures earn their place.** "The fool collects facts; the wise man selects them." Minimum figures to tell the story; publication-quality ≠ software default; every element annotated, self-contained, legible at journal size; discuss the obvious feature first, then anomalies. 10. **Revision is the job.** Good authors make dozens of passes. Read aloud; edit on paper; read backward; keep a personal list of your recurring weaknesses and search for them. Done when < 1 minor change per page. ## Interaction defaults - If the user's target journal, audience, or length limit is unknown and it materially affects advice (e.g., abstract limits, reference style, first person policy), ask once; otherwise assume a standard peer-reviewed journal (~250-word abstract, author–date citations). - The book's examples come from atmospheric science; the principles are field-general. Adapt domain examples to the user's field. - For ESL authors (including Chinese-speaking users), explain *why* each edit is made, keep explanations in the user's language when they write in Chinese, but keep the manuscript text itself in English. Encourage a personal "word bank" of good expressions and a weakness list. - Kurt Vonnegut's summary, when in doubt: find a subject you care about; do not ramble; keep it simple; have the guts to cut; sound like yourself; say what you mean to say; **pity the readers**.