Skip to content
SkillScript-led6,830 tokens in use

Huashu Design

Turns one sentence into a finished HTML design: prototypes, decks, animations, infographics and Xiaohongshu image sets, drafted as three real directions first.

29,964 stars on the repository MIT Licensealchaincyf/huashu-design
npx skills add alchaincyf/huashu-design
View source

Install command as written in the repository's README

Skill label

measured from its files

What installing Huashu Design costs

151tokens idle
6,830tokens in use
117,728tokens, up to

Idle is what sits in context while the skill is installed. In use is its instructions, loaded when it runs. Up to adds the files its instructions link. Counts are approximate, made offline with cl100k_base.

Model dependence Script-led

Script-ledMixedModel-led

Most of this skill is scripts, so it behaves much the same whichever model runs it.

This describes the skill's structure, how much of it is scripts. It does not say how well it works. Scripts: assets/deck_stage.js, assets/style-specimens/xhs/_kit/birds.js, assets/style-specimens/xhs/_kit/xhs.js, scripts/add-music.sh, scripts/build_specimens.py, scripts/capabilities.py, scripts/cloud/ai-review-video.py, scripts/cloud/tts-doubao.mjs, scripts/convert-formats.sh, scripts/design-gate-hook.sh, scripts/design_lint.py, scripts/export_deck_pdf.mjs, scripts/export_deck_pptx.mjs, scripts/export_deck_stage_pdf.mjs, scripts/fetch_images.py, scripts/gen_deck_thumbs.mjs, scripts/html2pptx.js, scripts/learn.py, scripts/mix-voiceover.sh, scripts/narrate-pipeline.mjs, scripts/polish.py, scripts/pptx_from_rendered.py, scripts/ref_board.py, scripts/render-narration.sh, scripts/render-video-seek.js, scripts/render-video.js, scripts/sfx-cues.sh, scripts/tests/test_compare_failures.py, scripts/tests/test_design_lint.py, scripts/tests/test_polish.py, scripts/tests/test_pptx_from_rendered.py, scripts/tests/test_pptx_svg_index.py, scripts/tests/test_xhs_render.py, scripts/verify-video.sh, scripts/verify.py, scripts/xhs_render.py.

Touches

shelldetectednetworkdetectedwrites filesdetectedcredentialsdetected

Declared tags come from the skill's frontmatter; detected ones from its scripts, its code blocks and the commands it names in inline code. This is detected, not audited.

How it scores

Quality

How well the skill is written, counted from its own files: its frontmatter, its steps, examples, stated output and constraints, its size and whether every file it ships is one it names.How we score it →
Good
  • Frontmatter8/15
  • Structure16/25
  • Examples12/15
  • Output7/15
  • Size11/15
  • Resources7/15

Security

Starts at 100 and loses points for each thing in the skill's files that a reader should look at first.How we score it →
Needs attention
  • File that was not read−15

    a file the skill ships that was not read, so what it does is unknown

    .env.example
  • Runs shell commands−5

    does this on the machine, and the frontmatter does not say so

    SKILL.md, line 67
  • Writes files−8

    does this on the machine, and the frontmatter does not say so

    SKILL.md, line 186
  • Minified code−10

    a line of 3605 characters: minified or generated, so it cannot be read

    assets/style-specimens/xhs/_kit/birds.js, line 2
  • Reaches the network−10

    does this on the machine, and the frontmatter does not say so

    scripts/build_specimens.py, line 78
  • Reads credentials−15

    does this on the machine, and the frontmatter does not say so

    scripts/cloud/ai-review-video.py, line 49

Repository health

Pre-computed score based on development activity, maintenance, community, maturity, and trend momentum.How we score it →
Good
  • Development activity88
  • Maintenance96
  • Community60
  • Maturity28
  • Momentum40

Huashu Design makes the agent work as a design studio that outputs HTML. For any new visual task it first shows three real draft directions chosen from six candidates, waits for your pick, and only then builds the full piece.

It covers web pages, slide decks, infographics, phone app prototypes, timeline animations and Xiaohongshu image sets, with a 92-style library, a brand asset protocol, a screenshot polish loop and export scripts for PDF, PPTX, MP4 and GIF. The instructions are written in Chinese.

What You Get

  • Single-file HTML prototypes, decks and infographics, with deck export to PDF and editable PPTX
  • Timeline animations exported as MP4 and GIF, with optional background music and narration
  • Three draft directions with thumbnails, plus three alternates, before any full build
  • A brand-spec file with real logos and colours when the design involves a named product
  • Local taste notes and a look ledger so later projects avoid repeating earlier designs

Common Use Cases

  • Prototype a clickable phone app with real photos and a device frame
  • Build a talk deck in the browser and export it as an editable PowerPoint
  • Make a product launch animation and render it to MP4 and GIF
  • Produce a set of 3:4 Xiaohongshu cards from one note
  • Run a six-dimension expert critique of an existing design

What Makes It Unique

The skill refuses to build until a human has seen real drafts. It generates candidates from three different logics (a best-designer view, a real reference captured as a screenshot, and a random pick from a style library), and it has a ledger that records how past projects looked so new ones avoid the same look. It also tells the agent about its own default taste, the warm cream and serif look, and asks it to design from the subject itself. Most design skills stop at a style guide; this one adds gate files, lint checks and a screenshot loop.

Under The Hood

The agent reads SKILL.md, picks a module handbook from references, and runs local Python and Node scripts for linting, screenshots, deck export and video rendering, using Playwright, ffmpeg and similar tools. It writes project files such as brand-spec, direction-approved and polish-log, and keeps taste notes and a look ledger in a folder in your home directory. It can search the web to verify facts, download brand logos and official assets, fetch Wikimedia images, and open reference pages in a headless browser for screenshots. Optional cloud scripts call ByteDance speech and video review APIs with your own keys, and only after an explicit confirmation flag. Once a month it can run a git command to check for a newer tagged version and writes a small date file. The agent can also install packages for its export tools.

Join founders building with open source

Opinionated takes, migration guides, cost-saving tips, and insights from the open source ecosystem.

Subscribe on SubstackJoin 750+ subscribers