# answer-with-html **Repository Path**: cjl2385/answer-with-html ## Basic Information - **Project Name**: answer-with-html - **Description**: No description available - **Primary Language**: Unknown - **License**: MIT - **Default Branch**: main - **Homepage**: None - **GVP Project**: No ## Statistics - **Stars**: 0 - **Forks**: 0 - **Created**: 2026-10-21 - **Last Updated**: 2026-11-16 ## Categories & Tags **Categories**: Uncategorized **Tags**: None ## README # Answer with HTML > Compile extended Markdown — yours or an AI's — into interactive HTML pages and step-by-step explainer videos, for technical explanations, option comparisons, code reviews, and shareable reports

English | 中文

MIT license Node.js
TypeScript Bun Vite
Tailwind CSS Mermaid Markdown

Write extended Markdown on the left, get an interactive HTML page on the right

## What it solves - **Long answers are hard to read** — content becomes multi-panel sheets or continuous docs, with flowcharts, sequence diagrams, ER diagrams, trees, charts, tables, and syntax-highlighted code - **Asking an AI to write a whole HTML page is unstable** — you author content and structure only; the compiler owns components, diagram coordinates, styles, and layout - **Answers are hard to keep and share** — output is static HTML with no app to deploy; local assets can be bundled alongside for offline reading - **Reading should lead to discussion** — pages offer multiple-choice questions and per-panel comments; the Reply action collects them as text you can paste back to the AI - **Charts stay readable without JavaScript** — `chart` fences compile to a static SVG with a legend and a native data table; `xy` covers line/area/bar/point/bubble/stacked/combo, alongside `pie`, `heatmap`, and `radar`. Pages add tooltips and keyboard navigation by default, while videos always use the static version - **Static explanations are not always enough** — the same diagram components can be rendered into explainer videos with scenes, narration, and camera focus - **Some ideas only click when you try them** — `tsx live` blocks run Preact components styled with Tailwind classes right in the page, with built-in UI primitives for forms, buttons, JSON and program output; `useApi` calls endpoints and, when the backend is down, falls back to a snapshot with an in-block explanation and a Retry button; code shared by a page's blocks goes in `tsx live-lib`, and `ts server` ships a backend with the document (requires Bun) ## Install as a skill for AI agents ```bash npx skills add https://gitee.com/cjl2385/answer-with-html.git ``` To verify locally, run `npx skills add /absolute/path/to/this/repo` in a scratch workspace: use the repo's real absolute path, not `.` pointing at the workspace; the install targets the current project and never overwrites existing skills. Once installed, the agent decides on its own when to compile an answer: when you ask for HTML, diagrams, or a visual explanation, when you say you don't understand and something needs to be explained step by step, or when you explicitly request an explainer video. It reads `SKILL.md` and calls the bundled CLI at `/scripts/to-html.mjs`. Compatibility: runs on Node.js 22.12+ or Bun ≥ 1.2. Media export additionally needs a browser (MP4 also ffmpeg). Image compression is built in when running under Bun; on Node, install sharp if you want it (raw bytes otherwise). Voice narration and the HTTP serve mode are covered in the skill references. ## Update the skill Check for updates, or update only this globally installed skill: ```bash npx skills check npx skills update answer-with-html -g ``` **Overwrite only the Claude Code global directory** ```bash npx skills add https://gitee.com/cjl2385/answer-with-html.git --skill answer-with-html --agent claude-code --global --copy --yes ``` ## Writing pages No file needed — pipe the draft through stdin (`-` reads it). Examples below shorten `node /scripts/to-html.mjs` to `to-html`: ````bash to-html render - <<'TO_HTML_EOF' --- title: How a request is handled template: sheet lang: en --- ## A Conclusion Plain Markdown: paragraphs, lists, tables, images, and code blocks all work. ## B Processing flow ```flow LR Input -> Parse -> Render -> HTML ``` TO_HTML_EOF ```` To render an existing draft, run `to-html render draft.md` (relative image paths resolve against the draft's directory; a stdin draft resolves against the current directory). `##` headings are panels; `sheet` suits multi-panel overviews, `doc` suits continuous reading. Diagrams use the project's own fence syntax — no hand-written SVG required. Without `-o`, output goes to a task directory `~/.answer-with-html/_/` (a timestamp is appended on name collisions): `_.origin.md` keeps the source draft (same name as the page), `_.html` (same name as its directory) is the page, local images are compressed into `assets/` by sharp or Bun's built-in engine (raw bytes only when neither is available); `TO_HTML_HOME` relocates the data root. The page opens in the browser automatically after rendering. ## Common operations ```bash # deliver together with images: generates assets/ next to the HTML plus a same-named .origin.md draft snapshot to-html render answer.md -o report/index.html # a single self-contained file: local images embedded as Base64 to-html render answer.md -o standalone.html --images embed # replace one panel of an existing page to-html patch report/index.html --panel A --from replacement.md # export real video files to-html video outline.md --voice off --mp4 to-html video outline.md --voice off --webm # syntax and options to-html help to-html help flow to-html help video ``` - In the default image mode, **the HTML file and its sibling `assets/` directory must be moved together**; in embed mode, a single local image is capped at 5 MB. `![caption](x.mp4)` / `.mp3` etc. render as players; local video and audio (including `