Relax default HTML allowlist

CHANGELOG.md

@@ -42,6 +42,7 @@
   - Bump markdown-it to [v14.1.0](https://github.com/markdown-it/markdown-it/blob/master/CHANGELOG.md#1410---2024-03-19), and follow the latest spec of [CommonMark 0.31.2](https://spec.commonmark.org/0.31.2/)
   - Support for CSS nesting (`cssNesting` constructor option)
 - Use simpler CSS minification when `minifyCSS` option is enabled ([#381](https://github.com/marp-team/marp-core/pull/381))
+- Relax HTML allowlist: Allowed a lot of HTML elements and attributes by default ([#301](https://github.com/marp-team/marp-core/issues/301), [#383](https://github.com/marp-team/marp-core/pull/383))
 
 * Upgrade development Node.js to v20 LTS ([#359](https://github.com/marp-team/marp-core/pull/359))
 * Upgrade dependent packages to the latest version ([#380](https://github.com/marp-team/marp-core/pull/380))

README.md

@@ -243,6 +243,7 @@ const marp = new Marp({
 
 Setting whether to render raw HTML in Markdown. It's an alias to `markdown.html` ([markdown-it option](https://markdown-it.github.io/markdown-it/#MarkdownIt.new)) but has additional feature about HTML allowlist.
 
+- (default): Use Marp's default allowlist.
 - `true`: The all HTML will be allowed.
 - `false`: All HTML except supported in Marpit Markdown will be disallowed.
 
@@ -265,7 +266,7 @@ By passing `object`, you can set the allowlist to specify allowed tags and attri
 }
 ```
 
-Marp core allows only `<br>` tag by default. That is defined in [a readonly `html` member in `Marp` class](https://github.com/marp-team/marp-core/blob/38fb33680c5837f9c48d8a88ac94b9f0862ab6c7/src/marp.ts#L34).
+By default, Marp Core allows known HTML elements and attributes that are considered as safe. That is defined as a readonly `html` member in `Marp` class. [See the full default allowlist in the source code.](src/html/allowlist.ts)
 
 > [!NOTE]
 > Whatever any option is selected, `<!-- HTML comment -->` and `<style>` tags are always parsed by Marpit for directives / tweaking style.

marp.config.mjs

@@ -7,7 +7,6 @@ export default {
     path.dirname(new URL(import.meta.url).pathname),
     './sandbox',
   ),
-  html: true,
   options: {
     minifyCSS: false,
   },

src/html/allowlist.ts

@@ -0,0 +1,242 @@
+export type HTMLAllowList = {
+  [tag: string]:
+    | string[]
+    | { [attr: string]: boolean | ((value: string) => string) }
+}
+
+const globalAttrs = {
+  class: true,
+  dir: (value) => {
+    const normalized = value.toLowerCase()
+    return ['rtl', 'ltr', 'auto'].includes(normalized) ? normalized : ''
+  },
+  lang: true,
+  title: true,
+} as const satisfies HTMLAllowList[string]
+
+const generateUrlSanitizer =
+  (schemas: string[]) =>
+  (value: string): string => {
+    if (value.includes(':')) {
+      // Check the URL schema if it exists
+      const trimmed = value.trim().toLowerCase()
+      const schema = trimmed.split(':', 1)[0]
+
+      for (const allowedSchema of schemas) {
+        if (schema === allowedSchema) return value
+        if (allowedSchema.includes(':') && trimmed.startsWith(allowedSchema))
+          return value
+      }
+
+      return ''
+    }
+    return value
+  }
+
+const webUrlSanitizer = generateUrlSanitizer(['http', 'https'])
+const imageUrlSanitizer = generateUrlSanitizer(['http', 'https', 'data:image/'])
+const srcSetSanitizer = (value: string): string => {
+  for (const src of value.split(',')) {
+    if (!imageUrlSanitizer(src)) return ''
+  }
+  return value
+}
+
+export const defaultHTMLAllowList = {
+  a: {
+    ...globalAttrs,
+    href: webUrlSanitizer,
+    name: true, // deprecated attribute, but still useful in Marp for making stable anchor link
+    rel: true,
+    target: true,
+  },
+  abbr: globalAttrs,
+  address: globalAttrs,
+  article: globalAttrs,
+  aside: globalAttrs,
+  audio: {
+    ...globalAttrs,
+    autoplay: true,
+    controls: true,
+    loop: true,
+    muted: true,
+    preload: true,
+    src: webUrlSanitizer,
+  },
+  b: globalAttrs,
+  bdi: globalAttrs,
+  bdo: globalAttrs,
+  big: globalAttrs,
+  blockquote: {
+    ...globalAttrs,
+    cite: webUrlSanitizer,
+  },
+  br: globalAttrs,
+  caption: globalAttrs,
+  center: globalAttrs, // deprecated
+  cite: globalAttrs,
+  code: globalAttrs,
+  col: {
+    ...globalAttrs,
+    align: true,
+    valign: true,
+    span: true,
+    width: true,
+  },
+  colgroup: {
+    ...globalAttrs,
+    align: true,
+    valign: true,
+    span: true,
+    width: true,
+  },
+  dd: globalAttrs,
+  del: {
+    ...globalAttrs,
+    cite: webUrlSanitizer,
+    datetime: true,
+  },
+  details: {
+    ...globalAttrs,
+    open: true,
+  },
+  div: globalAttrs,
+  dl: globalAttrs,
+  dt: globalAttrs,
+  em: globalAttrs,
+  figcaption: globalAttrs,
+  figure: globalAttrs,
+  // footer: globalAttrs, // Inserted by Marpit directives so disallowed to avoid confusion
+  h1: globalAttrs,
+  h2: globalAttrs,
+  h3: globalAttrs,
+  h4: globalAttrs,
+  h5: globalAttrs,
+  h6: globalAttrs,
+  // header: globalAttrs, // Inserted by Marpit directives so disallowed to avoid confusion
+  hr: globalAttrs,
+  i: globalAttrs,
+  img: {
+    ...globalAttrs,
+    align: true, // deprecated attribute, but still useful in Marp for aligning image
+    alt: true,
+    decoding: true,
+    height: true,
+    loading: true,
+    src: imageUrlSanitizer,
+    srcset: srcSetSanitizer,
+    title: true,
+    width: true,
+  },
+  ins: {
+    ...globalAttrs,
+    cite: webUrlSanitizer,
+    datetime: true,
+  },
+  kbd: globalAttrs,
+  li: {
+    ...globalAttrs,
+    type: true,
+    value: true,
+  },
+  mark: globalAttrs,
+  nav: globalAttrs,
+  ol: {
+    ...globalAttrs,
+    reversed: true,
+    start: true,
+    type: true,
+  },
+  p: globalAttrs,
+  picture: globalAttrs,
+  pre: globalAttrs,
+  source: {
+    height: true,
+    media: true,
+    sizes: true,
+    src: imageUrlSanitizer,
+    srcset: srcSetSanitizer,
+    type: true,
+    width: true,
+  },
+  q: {
+    ...globalAttrs,
+    cite: webUrlSanitizer,
+  },
+  rp: globalAttrs,
+  rt: globalAttrs,
+  ruby: globalAttrs,
+  s: globalAttrs,
+  section: globalAttrs,
+  small: globalAttrs,
+  span: globalAttrs,
+  sub: globalAttrs,
+  summary: globalAttrs,
+  sup: globalAttrs,
+  strong: globalAttrs,
+  strike: globalAttrs,
+  table: {
+    ...globalAttrs,
+    width: true,
+    border: true,
+    align: true,
+    valign: true,
+  },
+  tbody: {
+    ...globalAttrs,
+    align: true,
+    valign: true,
+  },
+  td: {
+    ...globalAttrs,
+    width: true,
+    rowspan: true,
+    colspan: true,
+    align: true,
+    valign: true,
+  },
+  tfoot: {
+    ...globalAttrs,
+    align: true,
+    valign: true,
+  },
+  th: {
+    ...globalAttrs,
+    width: true,
+    rowspan: true,
+    colspan: true,
+    align: true,
+    valign: true,
+  },
+  thead: {
+    ...globalAttrs,
+    align: true,
+    valign: true,
+  },
+  time: {
+    ...globalAttrs,
+    datetime: true,
+  },
+  tr: {
+    ...globalAttrs,
+    rowspan: true,
+    align: true,
+    valign: true,
+  },
+  u: globalAttrs,
+  ul: globalAttrs,
+  video: {
+    ...globalAttrs,
+    autoplay: true,
+    controls: true,
+    loop: true,
+    muted: true,
+    playsinline: true,
+    poster: imageUrlSanitizer,
+    preload: true,
+    src: webUrlSanitizer,
+    height: true,
+    width: true,
+  },
+  wbr: globalAttrs,
+} as const satisfies HTMLAllowList

src/html/html.ts

@@ -1,6 +1,5 @@
 import selfClosingTags from 'self-closing-tags'
-import { FilterXSS } from 'xss'
-import { friendlyAttrValue, escapeAttrValue } from 'xss/lib/default'
+import { FilterXSS, friendlyAttrValue, escapeAttrValue } from 'xss'
 import { MarpOptions } from '../marp'
 
 const selfClosingRegexp = /\s*\/?>$/
@@ -12,7 +11,7 @@ const xhtmlOutFilter = new FilterXSS({
     }
     return html
   },
-  whiteList: {},
+  allowList: {},
 })
 
 export function markdown(md): void {
@@ -60,7 +59,7 @@ export function markdown(md): void {
       }
 
       const filter = new FilterXSS({
-        whiteList: allowList,
+        allowList,
         onIgnoreTag: (_, rawHtml) => (html === true ? rawHtml : undefined),
         safeAttrValue: (tag, attr, value) => {
           let ret = friendlyAttrValue(value)

src/marp.ts

@@ -9,6 +9,7 @@ import * as autoScalingPlugin from './auto-scaling'
 import * as customElements from './custom-elements'
 import * as emojiPlugin from './emoji/emoji'
 import { generateHighlightJSInstance } from './highlightjs'
+import { defaultHTMLAllowList, type HTMLAllowList } from './html/allowlist'
 import * as htmlPlugin from './html/html'
 import * as mathPlugin from './math/math'
 import * as scriptPlugin from './script/script'
@@ -17,13 +18,7 @@ import * as slugPlugin from './slug/slug'
 
 export interface MarpOptions extends Options {
   emoji?: emojiPlugin.EmojiOptions
-  html?:
-    | boolean
-    | {
-        [tag: string]:
-          | string[]
-          | { [attr: string]: boolean | ((value: string) => string) }
-      }
+  html?: boolean | HTMLAllowList
   markdown?: object
   math?: mathPlugin.MathOptions
   minifyCSS?: boolean
@@ -36,7 +31,7 @@ export class Marp extends Marpit {
 
   private _highlightjs: HLJSApi | undefined
 
-  static readonly html = { br: [] }
+  static readonly html = defaultHTMLAllowList
 
   constructor(opts: MarpOptions = {}) {
     const mdOpts: Record<string, any> = {