name: shakespeare-writing-studio description: Produce publish-ready Chinese and English writing packages across WeChat articles, Medium/Substack essays, research notes, founder memos, op-eds, speeches, data storytelling, statistical graphics, and platform-ready HTML. Use this unified workflow for writing, rewriting, sourcing, lowering AI feel, creating visuals, and packaging final articles for Claude Code, Claude CLI, Codex CLI, and local agents.

Shakespeare Writing Studio

This workflow turns a rough idea, draft, notes, transcript, outline, or research question into a publish-ready writing package.

Treat Chinese and English work as modules inside one editorial system. Select modules by platform, language, article type, and visual need.

Main Deliverables

• final.md: publish-ready article, essay, memo, speech, or research note.
• platform HTML: wechat-ready.html, medium-ready.html, or clean web preview.
• visuals: assets/*.svg, charts/*.svg, cover, mechanism diagram, source-backed chart, statistical atlas, timeline, map, or data card.
• publishing-note.md: title options, subtitle/deck, social copy, source cautions, visual notes, AI-feel cleanup notes.

0. Select Modules

Always read the relevant modules before writing.

For every task:

• references/source-quality.md before any background search.
• references/human-voice.md before rewriting.
• references/editorial-lessons.md before drafting, titling, or designing visuals.
• references/visual-production.md before any cover, chart, diagram, or image.
• references/editorial-qa.md before delivery.

For Chinese WeChat:

• references/wechat/chinese-finance-tech-writing.md
• references/wechat/wechat-layout-patterns.md
• references/wechat/layout-wechat.md

For English long-form:

• references/medium-writing-patterns.md
• references/platform-formats.md

For research and data:

• references/research-data.md
• references/data-visuals.md

For final release:

• references/release-gate.md

1. Build The Spine

Extract:

• target language and platform
• intended reader
• core claim
• arguable stance
• title direction and search terms
• facts, dates, names, numbers
• source needs
• examples and scenes
• visual opportunities
• phrases that sound AI-generated

Before drafting, write the stance in one sentence. If it reads like "this trend is important," "值得关注," or "there are opportunities and challenges," rebuild the argument before polishing.

If the user gives a sharper title direction, preserve its tension and keywords. Do not smooth it into a safer generic title.

2. Choose The Format

Pick one primary format:

• Chinese WeChat finance / technology article
• English Medium / Substack essay
• research-backed article
• data-led report
• founder/operator note
• executive memo
• op-ed
• speech or keynote draft
• product / technology essay
• narrative argument

The format controls the opening, section rhythm, evidence density, visual density, and ending.

3. Research With Source Discipline

Use research only when it materially improves the piece.

Allowed:

• primary sources
• official announcements
• public filings
• reputable named reporting
• official datasets
• research institutions

Rejected:

• generic SEO pages
• unsourced blogs
• random statistics
• repost farms
• AI summaries

Every number used in the article or visual must have a source and date.

4. Write And Rewrite

Rules:

• Open with something concrete.
• Keep one central spine.
• Let paragraphs vary in length.
• Prefer concrete verbs over abstract nouns.
• Use examples before broad claims.
• Keep the writer's point of view visible.
• Remove generic AI scaffolding.
• Do not overbalance the argument into bland neutrality.
• Do not invent strategic frames that the sources and article do not support.

5. Visuals And Charts

Visuals are shared across Chinese and English work. Statistical graphics, covers, timelines, mechanism diagrams, and source-backed charts are available to both.

Before making a visual:

1. Write the visual job in one sentence.
2. Choose a direction from article content.
3. Pick the visual type.
4. Sketch 3–5 nodes, numbers, or cards.
5. Build the SVG/HTML.
6. Check readability at mobile width.

Hard rule: do not default to dark background. Dark visuals are allowed only when the article's subject, tone, and metaphor justify them.

For research-led articles, consider one hero statistical graphic when the reader needs to see a system rather than one metric.

Cover rule: the cover must be understood in 3 seconds on a phone. Avoid literal mountain / flag / summit graphics, generic route maps, cheap arrows, and abstract node diagrams unless the article itself demands that metaphor.

6. Platform Packaging

Produce:

• primary title
• alternate titles
• subtitle/deck
• platform-specific preview description
• social teaser
• suggested tags or WeChat summary
• source table if research was used
• visual list and source notes

7. QA

The piece is not done until:

• the stance is clear
• the opening earns attention
• each section advances the argument
• no unsupported claims remain
• no paragraph feels like generic AI filler
• data sources are listed when used
• visuals match the article direction
• charts clarify the argument
• cover does not lower the article's perceived quality
• platform preview reads well on mobile
• the ending lands with a thought, not a summary

Hard Rules

1. Do not invent facts, quotes, numbers, sources, or personal experiences.
2. Mark missing specifics as [TK] or ◻◻.
3. Background search must use top-tier sources only.
4. Remove obvious AI sentence scaffolds.
5. Never use visuals as decoration.
6. Never default to one visual style across projects.
7. Never use unsupported architecture/product language to make the article sound deeper.
8. If AI assistance materially shaped the article and the target platform requires disclosure, note it in publishing-note.md.

Shakespeare Writing Studio

Shakespeare Writing Studio is an agent-ready editorial system for producing publish-ready writing in Chinese and English.

It supports WeChat articles, Medium/Substack essays, research notes, op-eds, founder memos, speeches, data storytelling, statistical graphics, HTML previews, and complete publishing packages.

The goal is simple: turn a rough idea, draft, notes, transcript, or research question into work that has a point of view, reads like a human wrote it, cites serious sources, and looks ready to publish.

What It Produces

• final.md: publish-ready article, essay, memo, speech, or research note.
• Platform preview: wechat-ready.html, medium-ready.html, or clean web HTML.
• Visual assets: covers, timelines, evidence cards, mechanism diagrams, data charts, statistical atlases, maps, word clouds, or source-backed explainers.
• publishing-note.md: title options, subtitle/deck, social copy, source cautions, visual notes, paste steps, and AI-feel cleanup notes.

Core Standard

Strong writing needs more than fluent paragraphs.

This workflow checks for:

• a clear, arguable stance
• a title with tension and search value
• a concrete opening
• paragraph rhythm that works on mobile
• serious source discipline
• human sentence texture
• visuals that raise the article's perceived quality
• data graphics that clarify the argument
• platform-specific packaging
• a final QA gate before release

If a paragraph, chart, or cover could appear in any article on the internet, it usually needs to be rebuilt.

Examples

• Chinese WeChat article: Fable 5 被封禁，智谱冲万亿。中国模型登顶日即将来临？
• WeChat HTML preview
• English research article: The AI Boom Is Becoming A Power Grid Story
• Medium-style HTML preview
• Research charts and atlas

Workflow Modules

Start with:

• SKILL.md

Install As A Plugin

This repository is packaged for both marketplace-style plugin installation and direct agent reading.

Codex / OpenAI-style marketplace:

codex plugin marketplace add https://github.com/ZCZHAO-1999/shakespeare-writing-studio

Claude-style marketplace:

claude plugin marketplace add https://github.com/ZCZHAO-1999/shakespeare-writing-studio

If your agent does not support plugin marketplaces yet, point it at:

AGENTS.md

or copy this skill folder into your agent's skills directory:

plugins/shakespeare-writing-studio/skills/shakespeare-writing-studio

For chat-only tools such as DeepSeek, Doubao, Kimi, Qwen, or a browser chatbot, use the copy-paste prompt pack:

• DeepSeek / Doubao prompt pack

The chat-only path cannot install a plugin, but it can follow the same editorial workflow if you paste the prompt, then paste your topic, draft, notes, or source links.

Plugin manifests:

• Codex manifest
• Claude manifest
• Codex marketplace
• Claude marketplace

Core references:

• references/editorial-lessons.md: hard-won rules for titles, stance, anti-AI prose, WeChat rhythm, and cover quality.
• references/human-voice.md: human voice editing and AI-feel reduction.
• references/source-quality.md: top-tier source rules.
• references/research-data.md: research and evidence discipline.
• references/data-visuals.md: chart selection and statistical graphics.
• references/visual-production.md: visual direction and production rules.
• references/editorial-qa.md: final editorial checks.
• references/release-gate.md: go-live quality gate.

Platform references:

• references/wechat/chinese-finance-tech-writing.md: Chinese finance and technology article patterns.
• references/wechat/wechat-layout-patterns.md: WeChat mobile reading rhythm and layout archetypes.
• references/wechat/layout-wechat.md: WeChat-ready layout guidance.
• references/medium-writing-patterns.md: English Medium/Substack-style essays.
• references/platform-formats.md: platform-specific format choices.

Visual Principle

The visual style must come from the article.

No default dark mode. No default purple. No generic tech cards. No chart added just to look serious.

Before creating a visual, choose the job:

• cover
• proof
• explanation
• transition
• comparison
• statistical evidence
• share-card packaging

Then choose the direction from the story:

• Clean Analyst: finance, strategy, market analysis, public data
• Event Pricing Board: regulatory shock, market repricing, valuation narrative
• Energy / Infrastructure Report: power, climate, industrial systems, logistics
• Product Interface: SaaS, AI workflow, enterprise software
• Data Evidence: statistics, comparisons, source-backed claims
• Human Profile: founders, teams, interviews
• Conceptual Cover: abstract thesis when the metaphor is strong enough

Rejected cover patterns include literal mountain/flag/summit graphics, cheap up-arrows, generic node maps, unreadable tiny labels, decorative dark plates, and private metaphors the reader has to decode.

Source Standard

Background research must use serious sources.

Preferred:

• official company announcements and documentation
• regulators, agencies, standards bodies, and court records
• primary datasets and research institutions
• audited public filings
• reputable business, technology, and financial publications with named reporting

Rejected:

• generic SEO explainers
• unsourced statistics pages
• AI-summary pages
• random blogs with no primary trail
• reposts and screenshot-only evidence

Agent Compatibility

Works with:

• Claude Code
• Claude CLI
• Codex CLI
• local CLI agents that can read files, browse/search when needed, rewrite drafts, run scripts, and produce Markdown/HTML/SVG outputs

Search keywords:

• Shakespeare Writing Studio
• WeChat article workflow
• Medium writing workflow
• Claude Code writing skill
• Claude CLI writing workflow
• Codex CLI writing skill
• AI human voice editing
• data storytelling workflow
• research essay workflow
• statistical graphics workflow

Quality Gate

Before release:

• the central stance is visible
• the title has tension and search value
• no unsupported factual claims remain
• no weak source chain remains
• no generic AI filler remains
• no invented strategy/product frame remains
• no decorative chart remains
• no default visual style remains
• the cover improves the article's perceived quality
• HTML image paths work
• mobile layout is readable

Shakespeare Writing Studio 中文说明

Shakespeare Writing Studio 是一套面向 AI agent 的写作与出版工作流，用来生产可发布的中文和英文内容。

它可以处理微信公众号文章、Medium/Substack 英文长文、研究笔记、观点文章、创始人备忘录、演讲稿、数据叙事、统计图、信息图、HTML 预览和完整发布包。

目标很明确：把一个粗糙主题、草稿、访谈、资料包或研究问题，推进成一篇有观点、有人味、有来源、有排版、有视觉质量的成稿。

能交付什么

• final.md：可发布正文、长文、备忘录、演讲稿或研究笔记。
• 平台预览：wechat-ready.html、medium-ready.html 或干净网页 HTML。
• 视觉资产：封面、时间轴、证据卡、机制图、数据图、统计图谱、地图、词云或有来源的信息图。
• publishing-note.md：标题组、副标题、社交文案、来源提示、视觉说明、复制发布步骤和 AI 腔清理说明。

核心标准

顶级文章不只是句子流畅。

这套流程会检查：

• 有没有明确、可争论的观点
• 标题有没有张力和搜索价值
• 开头是否具体
• 手机阅读节奏是否顺
• 来源是否严肃
• 句子是否像真人写作
• 视觉是否抬高文章质感
• 数据图是否真的推进论证
• 是否适配目标平台
• 发布前是否通过质量门槛

如果一段话、一张图或一个封面可以放进任何文章里，它通常就需要重做。

示例

• 中文公众号文章：Fable 5 被封禁，智谱冲万亿。中国模型登顶日即将来临？
• 微信公众号 HTML 预览
• 英文研究文章：The AI Boom Is Becoming A Power Grid Story
• Medium 风格 HTML 预览
• 研究图表和统计图谱

工作流模块

从这里开始：

• SKILL.md

作为 Plugin 安装

这个仓库已经同时支持 marketplace 风格安装和普通 agent 直接读取。

Codex / OpenAI 风格 marketplace：

codex plugin marketplace add https://github.com/ZCZHAO-1999/shakespeare-writing-studio

Claude 风格 marketplace：

claude plugin marketplace add https://github.com/ZCZHAO-1999/shakespeare-writing-studio

如果你的 agent 还不支持 plugin marketplace，直接让它读取：

AGENTS.md

或者把这个 skill 文件夹复制到对应 agent 的 skills 目录：

plugins/shakespeare-writing-studio/skills/shakespeare-writing-studio

如果对方只会用 DeepSeek、豆包、Kimi、通义千问或网页聊天机器人，直接使用复制粘贴版 Prompt Pack：

• DeepSeek / 豆包 Prompt Pack

聊天工具无法真正安装 plugin，但只要先粘贴这段 prompt，再粘贴选题、草稿、资料或链接，就可以按同一套编辑流程工作。

插件文件：

• Codex manifest
• Claude manifest
• Codex marketplace
• Claude marketplace

核心参考：

• references/editorial-lessons.md：标题、观点、反 AI 腔、公众号节奏、封面质量的实战修正规则。
• references/human-voice.md：真人口吻和 AI 腔清理。
• references/source-quality.md：高质量来源规则。
• references/research-data.md：研究和证据规范。
• references/data-visuals.md：图表选择和统计图谱。
• references/visual-production.md：视觉方向和生产规范。
• [references/edi