cs-xiaohuang-skill
Codex skills for the Xiaohuang brand character: light hand-drawn slides, article covers and web slides in one style
- 76 ★ on the repository
- Last commit
- MIT
Minimal access
Why this level
- The skill only prepares images and pages, while generation runs through your agent model
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/cs-xiaohuang-skillDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewMinimal accessThe tool runs nothing and makes no network calls.
- The skill only prepares images and pages, while generation runs through your agent model
This is third-party code. Review the repository files before installing.
What it does
cs-xiaohuang-skill gives an agent a set of workflows around one brand character, Xiaohuang. The `cs-xiaohuang-skill` skill owns the character and static content: from a course outline, a talk topic or a long article it first produces a per-slide `slide blueprint`, then 5–12 16:9 PNG pages and a `contact sheet`. For articles it makes a 21:9 cover and three to six 16:9 illustrations with a `shot list`. The `cs-web-slides` skill builds a presentation as a web page: it first offers three real cover directions, and after a pick it delivers a 16:9 `index.html` with an `assets/` folder, keyboard and touch navigation, a progress bar and in-browser text editing. It also turns an existing PPTX into a web presentation. Product promo videos belong to a separate `cs-shotcraft-skill`, installed on its own. The character stays the same: it picks, connects, fixes and carries, helping the viewer see the key decision or relation on the page. Instructions and captions in the repo are in Chinese, with an English summary line.
Who it is for. Authors of courses, talks and blogs who prepare materials in Chinese and want one brand style for pages, covers and web slides with one character.
Good fit when
- A course outline must become 5–12 pages in one style with a contact sheet for approval
- A blog or newsletter article needs a 21:9 cover and three to six 16:9 illustrations
- A presentation should be shown in a browser and flipped by keys and touch
- A finished PPTX should become a web presentation shared by link
- You need a fixed character that does not change from page to page
Not a fit when
- A different character or brand: the look and rules are tied to Xiaohuang
- You do not read Chinese: skill docs and texts are in Chinese
- A promo video: a separate `cs-shotcraft-skill` handles it
Example request
Build pages in the Xiaohuang style from my lesson outline: a per-slide blueprint first, then PNGs and a contact sheet.Limitations
The skills target Codex: folders are copied into `${CODEX_HOME:-$HOME/.codex}/skills/`, with a Windows PowerShell variant of the commands. Video is made by a separate `cs-shotcraft-skill` installed on its own, and its base, the video-shotcraft library, lives in another repo. The character look and brand rules belong to the author, as `NOTICE.md` states. Image generation depends on the agent model, not on the skill. Documentation and captions are in Chinese.
How to disable. Delete the `cs-xiaohuang-skill` and `cs-web-slides` folders from the Codex skills directory. Generated pages and `assets/` folders stay in your projects and are removed by hand.
FAQ
How do I install the skills?
Clone the repo and copy the `cs-xiaohuang-skill` and `cs-web-slides` folders into `${CODEX_HOME:-$HOME/.codex}/skills/`.
What does the skill deliver by default?
For slides: a per-slide blueprint, 5–12 16:9 PNGs and a contact sheet. For articles: a 21:9 cover, three to six 16:9 illustrations and a shot list.
How do web slides work?
`cs-web-slides` offers three covers, then delivers a 16:9 `index.html` with keyboard and touch navigation, a progress bar and in-browser text editing.
Can a PPTX become a web presentation?
Yes, that is a `cs-web-slides` task: it converts a finished PPT or PPTX into a shareable page.
Where do promo videos come from?
A separate `cs-shotcraft-skill` makes the storyboard and a Remotion build with or without music. It is installed separately.
Discussion
No questions or reviews yet.
Related
Skillsxiaohu-ip-studio: article illustrations with a fixed character
xiaohu-ip-studio
A skill that illustrates Chinese articles: it decides which sections need a picture and draws them with a recurring character
A collection of open Agent Skills by Huashu for content work: slides, design, data processing, video scripts and text proofreading
A visual-director skill for Xiaohongshu carousels: style choice, page structure, image prompts and the post caption
Twelve Chinese-language skills: HTML pages, long-form writing, coding-agent PRDs, meeting minutes and WeChat publishing.
A set of 25 skills for Codex and Claude Code: articles, research, Remotion video, interfaces, GitHub and pre-release verification
A skill that generates and edits Xiaohongshu covers in 18 preset styles through Codex, Claude Code or the command line.
README in short
Covers three linked skills, four capabilities, install by copying into the Codex directory, an interactive demo and the character principles.
README badge
Original SKILL.md
--- name: cs-xiaohuang-skill description: "小黄(有温度)的统一入口:将中文文章、课程笔记、演讲提纲、产品说明、工作流或粗略想法,转成以小黄为叙事主角的 PPT-style 图像 deck,或生成文章的 21:9 封面与 16:9 正文配图;也生成、编辑和延展小黄的标准图、动作表情、三视图、2D/3D、联名和身份修复。用户提到小黄、有温度 IP、温度种子、爱心天线、小黄 PPT、文章配图、公众号配图、博客配图时使用。" metadata: author: "陈硕" collection: "CS Skills" source: "https://github.com/ChenShuo2004/cs-xiaohuang-skill" compatibility: "Codex and any agent that supports SKILL.md" --- # 小黄 IP、PPT 与文章配图 小黄是“有温度”的固定品牌角色。这个 skill 既能生成、编辑和延展小黄 IP 资产,也能把内容整理成小黄轻手绘 PPT-style 页面图,或一组服务于文章阅读的封面与正文配图。始终遵守:**可以换动作、场景、媒介和画法;不可以换掉角色身份。** ## 交付边界 - PPT 与文章配图均为完整 raster PNG 页面图;PPT 标准页面和文章正文图为 16:9,推荐 1920×1080。 - 文章或活动封面为 21:9,推荐 2520×1080;文章模式默认交付 1 张封面和 3–6 张正文图。 - 品牌 IP 延展可以交付标准图、动作 / 表情、三视图、2D / 3D、线稿、联名、场景插画或身份修复;根据用户指定尺寸和媒介生成,不强制使用 PPT 画布。 - 本 skill 中的“PPT / 幻灯片 / deck”均指最终视觉页面图,不是可编辑 PPTX、图片型 PPTX 或 PDF。 - 需要可交互 HTML、浏览器翻页、现场演示或把 PPTX 网页化时,转用 `cs-web-slides`;本 skill 不产出网页幻灯片。 - 不使用 HTML、SVG、canvas、脚本绘图或 `python-pptx` 作为主要视觉生成手段。用图像生成模型逐页生成;仅可在方向已确认后,对错误中文做确定性文字叠加。 - 复杂架构图、数据仪表盘、正式流程图、密集表格不适合本 skill;应转用可编辑演示文稿或图表工作流。 ## 先读什么 所有生产任务都必须查看 `assets/reference-xiaohuang.png`,它是小黄的唯一身份锚点。按当前任务按需加载: - `references/intake.md`:输入判断、默认值与何时追问。 - `references/narrative-planning.md`:选择叙事骨架和生成 slide blueprint。 - `references/slide-archetypes.md`:按内容语义选择页面结构。 - `references/article-illustrations.md`:文章封面、正文 shot list 与阅读场景的构图规则。 - `references/character-dna.md`:小黄不可改变的形态与颜色规则。 - `references/media-and-variants.md`:2D / 3D、黑白、动作、联名与媒介转换规则。 - `references/ip-prompt-templates.md`:品牌 IP 标准图、3D、风格迁移、联名与局部修复提示词。 - `references/visual-dna.md`:整套页面的构图、文字、色彩和角色规则。 - `references/prompt-patterns.md`:多页 style lock 与逐页图像提示词。 - `references/output-quality.md`:最终交付和质量验收。 使用 `assets/theme-tokens.json` 写紧凑的风格锁定。当前图像工具支持本地参考图时,必须把小黄参考图作为身份参考传入;不支持时,使用角色 DNA 的文字锚点,并在交付中说明是 prompt-matched。 ## 工作流 1. **读懂素材。** 读取文章、Markdown、PDF、DOCX、既有 PPTX、课程大纲、讲稿或粗略想法。既有 PPTX 只提取内容与视觉意图,不在本 skill 内编辑或包装它。 2. **做 intake。** 判断主题、受众、场景、交付长度、核心论点、可视化关系与证据充分度。只有缺少会改变结构的关键信息时,才问 1–3 个简短问题。 3. **选择输出模式。** 用户要标准图、动作、3D、联名或修复时选择 **brand IP**,并读取 `references/media-and-variants.md` 和 `references/ip-prompt-templates.md`;用户要静态图片课件、演讲页或演示图时选择 **deck**;用户要文章、公众号、博客或 Newsletter 配图时选择 **article illustration**。用户要可交互网页幻灯片、浏览器翻页或把 PPTX 网页化时,转用 `cs-web-slides`,不要在本 skill 内做成 PNG deck。文章模式先给 shot list;deck 模式先定叙事。内容模式中每页 / 每张图只能有一个主观点。 4. **语义映射版式。** 读取 `references/slide-archetypes.md`。用“对比、转化、筛选、阻塞、分叉、循环、搭建、总结”等页面角色建立节奏,不能机械重复一个模板。 5. **锁定整套风格。** 读取 `references/character-dna.md`、`references/visual-dna.md`,先写一次 style lock,再逐张复用。固定画布、暖白底、线条、调色、留白和小黄身份;deck 额外固定标题与页码,文章正文图默认不放这两个元素。 6. **先出 blueprint 或 shot list。** 用户要求规划、素材不足,或 deck 超过 5 页时,先给 slide-by-slide blueprint。文章配图无论是否生成,都先给 shot list:插入位置、核心意思、构图类型、小黄动作、物件和短标注。 7. **逐张生图。** 每个页面或配图单独调用图像工具;每张只表达一个关系。把小黄参考图传入并用 `references/prompt-patterns.md` 组装提示词。每张使用 `Required text only`,限制可见中文。