提示词库

Writing & Academic:Agent 系统提示词。

awesome-prompts 里维护的角色提示词和系统提示词,按领域分组。复制到 Claude Code、Codex 或任意对话模型里就能用。

15 条写作类系统提示词:去掉 AI 腔的编辑、技术写作和科研写作,以及学术论文的审稿和校对。把它放进系统提示词,再发你的文字。下面挑了 4 条,写了我们发的消息和它的回复,2026 年 11 月 15 日用 Claude Haiku 5.5 实测。

15 Writing & Academic 提示词

Academic Paper Architect EN

Full-spectrum manuscript orchestrator: 12-agent pipeline (literature strategy → structure → argument → draft → citation → bilingual abstract → simulated peer review → formatting); style calibration, writing quality checks, IRON RULE checkpoints, 8 invocation modes; based on Imbad0202/academic-research-skills (May 2026, 18k+ stars)

Academic Peer Reviewer EN

Comprehensive manuscript review: contribution assessment, methodology critique, reproducibility, ethics, constructive feedback, recommendation with confidence (2026)

Agent Style Enforcer EN

Literature-backed technical-prose writing ruleset: 21 rules (12 canonical from Strunk & White/Orwell/Pinker/Gopen & Swan + 9 field-observed from LLM output 2022, 2026) with severity tiers, BAD/GOOD examples, and escape hatch; drop-in for any AI agent producing `.md`, `.tex`, `.rst`, or source-code comments; based on yzhao062/agent-style (2026)

Humanizer EN

Writing editor that removes 29 signs of AI-generated text: detects inflated symbolism, promotional language, vague attributions, AI vocabulary, passive voice, filler phrases; supports voice calibration via writing samples; dual-pass audit workflow; based on blader/humanizer (Jan 2026)

Journal Adapt Writing Architect EN

Dynamic, corpus-grounded academic writing skill generator: learns target-journal conventions from user-provided papers, builds a reviewable `dynamic_writing_skill.md`, then revises manuscripts section by section with a 5-layer priority system (hard preserve → target journal → secondary corpus → static base → cleanup); based on WantongC/journal-adapt-writing-skill (May 2026, 438 stars)

LaTeX Academic Expert EN

Venue-aware LaTeX formatting + academic writing polish: template switching (NeurIPS/ICML/CVPR/ACL/IEEE/Nature/Science), citation-style conversion, page-limit compliance, double-blind anonymization, section-aware prose editing, Chinglish pattern fixes; preserves all commands/math/cites; based on Calix-L/awesome-latex-skills (May 2026, 171 stars)

Literature Professor EN

Essay writing and literary analysis from a professor's perspective

Nature-Style Scientific Writer EN

Submission-grade scientific writing and figure architect for Nature-family journals: argument-first drafting, hourglass structure, section-specific templates (abstract/introduction/results/discussion), verb calibration, publication-quality Python/R figure pipelines, data-availability ethics, and Chinese-author support; based on Yuan1z0825/nature-skills (Apr 2026, 7.3k+ stars)

Paper Figure Mirror Engineer EN

Camera-ready matplotlib figure architect: transfers the visual style of a top-conference paper figure (NeurIPS/ICML/ICLR/Nature) onto the user's data via iterative Drawer/Reviewer loops; enforces layout invariants (no overlap, no clipping, no defaults), L1-reference + L2-convention dual anchoring, and visible-but-recessive hairline calibration; outputs self-contained `.py` + camera-ready PDF/PNG; based on VILA-Lab/FigMirror (May 2026, 427 stars)

Paper Spine Architect EN

Motivation-driven academic paper mastery: motivation spine extraction, central argument trees, evidence-aware blueprints, revision matrices with argument-impact gating, and LaTeX-safe audits; based on WUBING2023/PaperSpine (May 2026, 1.7k+ stars)

Research Paper Proofreader EN

Claude Code/Codex paper proofreading: two-phase detect-then-fix workflow, 9 review categories (language, clarity, structure, LaTeX, notation), severity-graded issues, anti-AI-slop rules; based on LimHyungTae/awesome-claudecode-paper-proofreading (Mar 2026)

Simplified Technical English (STE) Writer EN

Agent skill that writes docs in ASD-STE100 Simplified Technical English: 20/25-word sentence limits, one word one meaning, simple tenses, active voice, condition-before-command; 72.9% fewer STE violations measured across 6 Claude models; based on AminBlg/SimpleEnglish (MIT, 1.7k+ stars, July 2026)

Stop-Slop Writing Editor EN

Prose editor that strips predictable AI tells: active voice, no adverbs, no throat-clearing, no binary contrasts, no em dashes; 5-dimension scorecard (directness, rhythm, trust, authenticity, density) with 35/50 revision threshold; based on hardikpandya/stop-slop (2026, 10.3k stars)

Talk-Normal Enabler EN

System prompt that removes AI slop: direct, informative, no filler/fluff/summary-stamps, no negation-based contrastive phrasing; 72, 73% token reduction on GPT-4o-mini/GPT-5.4 with zero information loss; based on hexiecs/talk-normal (2026)

Technical Writer EN

Senior dev-docs writer: Stripe/Twilio/Google standards; blog posts, API docs, release notes, READMEs; no padding (2026)

怎么用这几条提示词,附实测回复

Humanizer:去掉营销腔

我们给它一句话,讲在当今快节奏的数字环境里有个开创性平台。它先写一版平实的草稿,指出哪里还像机器写的,再改成一句具体的话,列出删掉的每个说法,最后问产品到底做什么。 来源:Humanizer

Technical Writer:能照着跑的快速上手

为一个删除无用 import 的命令行工具,它先写明面向的读者水平,再写 pipx 安装、一个带两个无用 import 的示例文件、运行命令和运行后的文件,最后列出发布前要核实的事项,第一项是包名。 来源:Technical Writer

Academic Peer Reviewer:审一段摘要

我们给它一段摘要:用一家医院的 500 张胸片训练,肺炎识别准确率 99%,证明 AI 能取代放射科医生。它先说明只看到了摘要,建议当前形式拒稿,并解释单一医院、缺少对照、结论缺乏支撑,让这个数字无从解读。 来源:Academic Peer Reviewer

Research Paper Proofreader:一段话,逐条数错

一段有主谓搭配错误和夸大结论的文字,它判为需要修改,数出 1 个严重、1 个主要、6 个次要和 1 个风格问题,每条附改法,从 it's 改 its、results shows 改 show,到一句写明指标并引用结果表的改写。 来源:Research Paper Proofreader