@aquera/nile-elements 1.9.7 → 1.9.9

package/dist/src/nile-wysiwyg-editor/engine/sanitize.js

@@ -5,10 +5,10 @@
  * string ever touches a live DOM, and strips style declarations down to the
  * set the schema understands.
  */
-import { isAllowedEmbedSrc } from './plugins/embeds';
+import { isAllowedEmbedSrc, EMBED_SANDBOX, EMBED_REFERRER_POLICY } from './plugins/embeds';
 /** Elements removed together with their content. */
 const DROP_WITH_CONTENT = new Set([
-    'script', 'style', 'frame', 'frameset', 'object', 'embed',
+    'script', 'frame', 'frameset', 'object', 'embed',
     'applet', 'form', 'input', 'textarea', 'select', 'button', 'meta', 'link',
     'base', 'title', 'noscript', 'template', 'svg', 'math', 'canvas', 'audio',
     'video', 'dialog', 'slot',
@@ -39,9 +39,21 @@ const TAG_ATTRS = {
     th: new Set(['colspan', 'rowspan', 'data-colwidth']),
     col: new Set(['span', 'width']),
     div: new Set(['data-embed-provider', 'data-width', 'data-align']),
-    iframe: new Set(['src', 'title', 'frameborder', 'allowfullscreen', 'allow']),
+    iframe: new Set([
+        'src', 'title', 'frameborder', 'allowfullscreen', 'allow', 'sandbox', 'referrerpolicy',
+    ]),
 };
-const GLOBAL_ATTRS = new Set(['style', 'align']);
+const GLOBAL_ATTRS = new Set(['style', 'align', 'class', 'id']);
+/** Author class/id names starting with these are the editor's own; drop them so
+ * content can't impersonate or restyle the chrome. */
+const RESERVED_NAME = /^(?:nile-wysiwyg|ProseMirror)/;
+/** Filtered `class`: reserved-prefixed tokens removed. Empty → drop attribute. */
+function cleanClassAttr(value) {
+    return value
+        .split(/\s+/)
+        .filter(token => token && !RESERVED_NAME.test(token))
+        .join(' ');
+}
 /** Style declarations the schema can represent. Everything else is noise. */
 const ALLOWED_STYLES = new Set([
     'color', 'background-color', 'font-family', 'font-size', 'font-style',
@@ -89,6 +101,34 @@ function filterStyle(value, tag) {
     })
         .join('; ');
 }
+/**
+ * Sanitizes the CSS inside a <style> block. Neutralizes the vectors that make
+ * a raw stylesheet dangerous while leaving ordinary rules intact:
+ *  - `@import` (external fetch / data exfiltration) — removed,
+ *  - `expression()`, `behavior:`, `-moz-binding:` (script execution) — broken,
+ *  - `url(...)` pointing anywhere other than a safe data: image or a local
+ *    `#fragment` — replaced with `url(#)`,
+ *  - any `</style` sequence — stripped so the text can't break out of the
+ *    element when the document is re-serialized.
+ * CSS comments are dropped first so none of the above can hide inside one.
+ */
+export function sanitizeCss(css) {
+    return css
+        .replace(/<\/style/gi, '')
+        .replace(/\/\*[\s\S]*?\*\//g, '')
+        .replace(/@import[^;]*;?/gi, '')
+        .replace(/expression\s*\(/gi, 'blocked(')
+        .replace(/(behavior|-moz-binding)(\s*):/gi, 'blocked-prop$2:')
+        .replace(/url\(\s*(['"]?)([^)]*?)\1\s*\)/gi, (full, _quote, raw) => {
+        const url = String(raw).trim();
+        if (/^#/.test(url))
+            return full;
+        if (/^data:image\/(png|gif|jpe?g|webp|avif|bmp);/i.test(url))
+            return full;
+        return 'url(#)';
+    })
+        .trim();
+}
 function sanitizeElement(el) {
     const tag = el.tagName.toLowerCase();
     const allowed = TAG_ATTRS[tag];
@@ -107,6 +147,17 @@ function sanitizeElement(el) {
             else
                 el.removeAttribute('style');
         }
+        else if (name === 'class') {
+            const cleaned = cleanClassAttr(attr.value);
+            if (cleaned)
+                el.setAttribute('class', cleaned);
+            else
+                el.removeAttribute('class');
+        }
+        else if (name === 'id') {
+            if (RESERVED_NAME.test(attr.value))
+                el.removeAttribute('id');
+        }
         else if (name === 'href' || name === 'src') {
             // Raster data: URLs are allowed on images only.
             if (!isSafeUrl(attr.value, tag === 'img'))
@@ -127,9 +178,23 @@ function walk(node) {
             child.remove();
             continue;
         }
-        // iframes are allowed ONLY for allowlisted embed providers.
-        if (tag === 'iframe' && !isAllowedEmbedSrc(child.getAttribute('src') ?? '')) {
-            child.remove();
+        // iframes are allowed for any https src, but always force-sandboxed so an
+        // untrusted embed can never navigate the top window or run unrestricted.
+        if (tag === 'iframe') {
+            if (!isAllowedEmbedSrc(child.getAttribute('src') ?? '')) {
+                child.remove();
+                continue;
+            }
+            child.setAttribute('sandbox', EMBED_SANDBOX);
+            child.setAttribute('referrerpolicy', EMBED_REFERRER_POLICY);
+        }
+        // <style> blocks are kept, but their CSS is sanitized and any attributes
+        // (media, nonce, …) dropped. Handled before the unwrap branch below so the
+        // raw CSS is never spilled into the document as visible text.
+        if (tag === 'style') {
+            child.textContent = sanitizeCss(child.textContent ?? '');
+            for (const attr of Array.from(child.attributes))
+                child.removeAttribute(attr.name);
             continue;
         }
         if (!ALLOWED_TAGS.has(tag)) {
@@ -148,7 +213,10 @@ function walk(node) {
  * DOMParser, so embedded scripts never execute and images never load.
  */
 export function sanitizeHtml(html) {
-    const doc = new DOMParser().parseFromString(html, 'text/html');
+    // Wrap in an explicit <body> so metadata-like tags (notably <style>) stay in
+    // the body instead of being hoisted into <head> by the HTML parser — which
+    // would drop them before walk() ever sees them. Mirrors htmlToDoc's parse.
+    const doc = new DOMParser().parseFromString(`<body>${html}</body>`, 'text/html');
     walk(doc.body);
     return doc.body.innerHTML;
 }

package/dist/src/nile-wysiwyg-editor/engine/sanitize.js.map

@@ -1 +1 @@
-{"version":3,"file":"sanitize.js","sourceRoot":"","sources":["../../../../src/nile-wysiwyg-editor/engine/sanitize.ts"],"names":[],"mappings":"AAAA;;;;;;GAMG;AAEH,OAAO,EAAE,iBAAiB,EAAE,MAAM,kBAAkB,CAAC;AAErD,oDAAoD;AACpD,MAAM,iBAAiB,GAAG,IAAI,GAAG,CAAC;IAChC,QAAQ,EAAE,OAAO,EAAE,OAAO,EAAE,UAAU,EAAE,QAAQ,EAAE,OAAO;IACzD,QAAQ,EAAE,MAAM,EAAE,OAAO,EAAE,UAAU,EAAE,QAAQ,EAAE,QAAQ,EAAE,MAAM,EAAE,MAAM;IACzE,MAAM,EAAE,OAAO,EAAE,UAAU,EAAE,UAAU,EAAE,KAAK,EAAE,MAAM,EAAE,QAAQ,EAAE,OAAO;IACzE,OAAO,EAAE,QAAQ,EAAE,MAAM;CAC1B,CAAC,CAAC;AAEH,gFAAgF;AAChF,MAAM,YAAY,GAAG,IAAI,GAAG,CAAC;IAC3B,GAAG,EAAE,IAAI,EAAE,IAAI,EAAE,IAAI,EAAE,IAAI,EAAE,IAAI,EAAE,IAAI,EAAE,YAAY,EAAE,IAAI,EAAE,IAAI,EAAE,IAAI;IACvE,IAAI,EAAE,IAAI,EAAE,GAAG,EAAE,GAAG,EAAE,QAAQ,EAAE,GAAG,EAAE,IAAI,EAAE,GAAG,EAAE,GAAG,EAAE,QAAQ,EAAE,KAAK;IACpE,MAAM,EAAE,MAAM,EAAE,KAAK,EAAE,KAAK,EAAE,KAAK,EAAE,MAAM;IAC3C,KAAK,EAAE,QAAQ,EAAE,YAAY,EAAE,KAAK,EAAE,QAAQ;IAC9C,OAAO,EAAE,OAAO,EAAE,OAAO,EAAE,OAAO,EAAE,IAAI,EAAE,IAAI,EAAE,IAAI,EAAE,UAAU,EAAE,KAAK;CACxE,CAAC,CAAC;AAEH,2DAA2D;AAC3D,MAAM,SAAS,GAAgC;IAC7C,CAAC,EAAE,IAAI,GAAG,CAAC,CAAC,MAAM,EAAE,OAAO,EAAE,QAAQ,EAAE,KAAK,EAAE,IAAI,EAAE,MAAM,EAAE,eAAe,CAAC,CAAC;IAC7E,KAAK,EAAE,IAAI,GAAG,CAAC,CAAC,MAAM,EAAE,aAAa,CAAC,CAAC;IACvC,EAAE,EAAE,IAAI,GAAG,CAAC,CAAC,OAAO,EAAE,MAAM,EAAE,iBAAiB,CAAC,CAAC;IACjD,EAAE,EAAE,IAAI,GAAG,CAAC,CAAC,gBAAgB,CAAC,CAAC;IAC/B,MAAM,EAAE,IAAI,GAAG,CAAC,CAAC,YAAY,CAAC,CAAC;IAC/B,EAAE,EAAE,IAAI,GAAG,CAAC,CAAC,gBAAgB,EAAE,cAAc,CAAC,CAAC;IAC/C,IAAI,EAAE,IAAI,GAAG,CAAC;QACZ,sBAAsB,EAAE,kBAAkB,EAAE,oBAAoB;QAChE,kBAAkB,EAAE,kBAAkB,EAAE,WAAW,EAAE,mBAAmB;KACzE,CAAC;IACF,GAAG,EAAE,IAAI,GAAG,CAAC,CAAC,eAAe,CAAC,CAAC;IAC/B,GAAG,EAAE,IAAI,GAAG,CAAC,CAAC,KAAK,EAAE,KAAK,EAAE,OAAO,EAAE,OAAO,EAAE,QAAQ,CAAC,CAAC;IACxD,EAAE,EAAE,IAAI,GAAG,CAAC,CAAC,SAAS,EAAE,SAAS,EAAE,eAAe,CAAC,CAAC;IACpD,EAAE,EAAE,IAAI,GAAG,CAAC,CAAC,SAAS,EAAE,SAAS,EAAE,eAAe,CAAC,CAAC;IACpD,GAAG,EAAE,IAAI,GAAG,CAAC,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;IAC/B,GAAG,EAAE,IAAI,GAAG,CAAC,CAAC,qBAAqB,EAAE,YAAY,EAAE,YAAY,CAAC,CAAC;IACjE,MAAM,EAAE,IAAI,GAAG,CAAC,CAAC,KAAK,EAAE,OAAO,EAAE,aAAa,EAAE,iBAAiB,EAAE,OAAO,CAAC,CAAC;CAC7E,CAAC;AAEF,MAAM,YAAY,GAAG,IAAI,GAAG,CAAC,CAAC,OAAO,EAAE,OAAO,CAAC,CAAC,CAAC;AAEjD,6EAA6E;AAC7E,MAAM,cAAc,GAAG,IAAI,GAAG,CAAC;IAC7B,OAAO,EAAE,kBAAkB,EAAE,aAAa,EAAE,WAAW,EAAE,YAAY;IACrE,aAAa,EAAE,YAAY,EAAE,iBAAiB,EAAE,sBAAsB;IACtE,gBAAgB,EAAE,aAAa,EAAE,qBAAqB,EAAE,iBAAiB;IACzE,aAAa;CACd,CAAC,CAAC;AAEH,MAAM,kBAAkB,GAAG,CAAC,OAAO,EAAE,QAAQ,EAAE,SAAS,EAAE,MAAM,EAAE,MAAM,CAAC,CAAC;AAE1E,MAAM,UAAU,SAAS,CAAC,GAAW,EAAE,cAAc,GAAG,KAAK;IAC3D,MAAM,OAAO,GAAG,GAAG,CAAC,IAAI,EAAE,CAAC;IAC3B,IAAI,OAAO,KAAK,EAAE;QAAE,OAAO,KAAK,CAAC;IACjC,uDAAuD;IACvD,IAAI,qBAAqB,CAAC,IAAI,CAAC,OAAO,CAAC;QAAE,OAAO,IAAI,CAAC;IACrD,MAAM,KAAK,GAAG,wBAAwB,CAAC,IAAI,CAAC,OAAO,CAAC,CAAC;IACrD,IAAI,CAAC,KAAK;QAAE,OAAO,IAAI,CAAC;IACxB,MAAM,QAAQ,GAAG,GAAG,KAAK,CAAC,CAAC,CAAC,CAAC,WAAW,EAAE,GAAG,CAAC;IAC9C,IAAI,kBAAkB,CAAC,QAAQ,CAAC,QAAQ,CAAC;QAAE,OAAO,IAAI,CAAC;IACvD,IAAI,cAAc,IAAI,QAAQ,KAAK,OAAO,EAAE,CAAC;QAC3C,6DAA6D;QAC7D,OAAO,8CAA8C,CAAC,IAAI,CAAC,OAAO,CAAC,CAAC;IACtE,CAAC;IACD,OAAO,KAAK,CAAC;AACf,CAAC;AAED,0EAA0E;AAC1E,MAAM,UAAU,GAAgC;IAC9C,GAAG,EAAE,IAAI,GAAG,CAAC,CAAC,OAAO,EAAE,QAAQ,CAAC,CAAC;IACjC,MAAM,EAAE,IAAI,GAAG,CAAC,CAAC,OAAO,EAAE,cAAc,CAAC,CAAC;IAC1C,GAAG,EAAE,IAAI,GAAG,CAAC,CAAC,OAAO,EAAE,WAAW,CAAC,CAAC;CACrC,CAAC;AAEF,SAAS,WAAW,CAAC,KAAa,EAAE,GAAW;IAC7C,MAAM,KAAK,GAAG,UAAU,CAAC,GAAG,CAAC,CAAC;IAC9B,OAAO,KAAK;SACT,KAAK,CAAC,GAAG,CAAC;SACV,GAAG,CAAC,IAAI,CAAC,EAAE,CAAC,IAAI,CAAC,IAAI,EAAE,CAAC;SACxB,MAAM,CAAC,IAAI,CAAC,EAAE;QACb,MAAM,IAAI,GAAG,IAAI,CAAC,KAAK,CAAC,CAAC,EAAE,IAAI,CAAC,OAAO,CAAC,GAAG,CAAC,CAAC,CAAC,IAAI,EAAE,CAAC,WAAW,EAAE,CAAC;QACnE,IAAI,CAAC,cAAc,CAAC,GAAG,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC,KAAK,IAAI,KAAK,CAAC,GAAG,CAAC,IAAI,CAAC,CAAC;YAAE,OAAO,KAAK,CAAC;QAC3E,sDAAsD;QACtD,OAAO,CAAC,2BAA2B,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;IACjD,CAAC,CAAC;SACD,IAAI,CAAC,IAAI,CAAC,CAAC;AAChB,CAAC;AAED,SAAS,eAAe,CAAC,EAAW;IAClC,MAAM,GAAG,GAAG,EAAE,CAAC,OAAO,CAAC,WAAW,EAAE,CAAC;IACrC,MAAM,OAAO,GAAG,SAAS,CAAC,GAAG,CAAC,CAAC;IAC/B,KAAK,MAAM,IAAI,IAAI,KAAK,CAAC,IAAI,CAAC,EAAE,CAAC,UAAU,CAAC,EAAE,CAAC;QAC7C,MAAM,IAAI,GAAG,IAAI,CAAC,IAAI,CAAC,WAAW,EAAE,CAAC;QACrC,MAAM,IAAI,GACR,CAAC,YAAY,CAAC,GAAG,CAAC,IAAI,CAAC,IAAI,CAAC,OAAO,KAAK,SAAS,IAAI,OAAO,CAAC,GAAG,CAAC,IAAI,CAAC,CAAC,CAAC;YACxE,CAAC,IAAI,CAAC,UAAU,CAAC,IAAI,CAAC,CAAC;QACzB,IAAI,CAAC,IAAI,EAAE,CAAC;YACV,EAAE,CAAC,eAAe,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;YAC9B,SAAS;QACX,CAAC;QACD,IAAI,IAAI,KAAK,OAAO,EAAE,CAAC;YACrB,MAAM,QAAQ,GAAG,WAAW,CAAC,IAAI,CAAC,KAAK,EAAE,GAAG,CAAC,CAAC;YAC9C,IAAI,QAAQ;gBAAE,EAAE,CAAC,YAAY,CAAC,OAAO,EAAE,QAAQ,CAAC,CAAC;;gBAC5C,EAAE,CAAC,eAAe,CAAC,OAAO,CAAC,CAAC;QACnC,CAAC;aAAM,IAAI,IAAI,KAAK,MAAM,IAAI,IAAI,KAAK,KAAK,EAAE,CAAC;YAC7C,gDAAgD;YAChD,IAAI,CAAC,SAAS,CAAC,IAAI,CAAC,KAAK,EAAE,GAAG,KAAK,KAAK,CAAC;gBAAE,EAAE,CAAC,eAAe,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;QAC3E,CAAC;aAAM,IAAI,IAAI,KAAK,QAAQ,EAAE,CAAC;YAC7B,IAAI,IAAI,CAAC,KAAK,KAAK,QAAQ;gBAAE,EAAE,CAAC,YAAY,CAAC,KAAK,EAAE,qBAAqB,CAAC,CAAC;;gBACtE,EAAE,CAAC,eAAe,CAAC,QAAQ,CAAC,CAAC;QACpC,CAAC;IACH,CAAC;AACH,CAAC;AAED,SAAS,IAAI,CAAC,IAAa;IACzB,KAAK,MAAM,KAAK,IAAI,KAAK,CAAC,IAAI,CAAC,IAAI,CAAC,QAAQ,CAAC,EAAE,CAAC;QAC9C,MAAM,GAAG,GAAG,KAAK,CAAC,OAAO,CAAC,WAAW,EAAE,CAAC;QACxC,IAAI,iBAAiB,CAAC,GAAG,CAAC,GAAG,CAAC,EAAE,CAAC;YAC/B,KAAK,CAAC,MAAM,EAAE,CAAC;YACf,SAAS;QACX,CAAC;QACD,4DAA4D;QAC5D,IAAI,GAAG,KAAK,QAAQ,IAAI,CAAC,iBAAiB,CAAC,KAAK,CAAC,YAAY,CAAC,KAAK,CAAC,IAAI,EAAE,CAAC,EAAE,CAAC;YAC5E,KAAK,CAAC,MAAM,EAAE,CAAC;YACf,SAAS;QACX,CAAC;QACD,IAAI,CAAC,YAAY,CAAC,GAAG,CAAC,GAAG,CAAC,EAAE,CAAC;YAC3B,iEAAiE;YACjE,oCAAoC;YACpC,IAAI,CAAC,KAAK,CAAC,CAAC;YACZ,KAAK,CAAC,WAAW,CAAC,GAAG,KAAK,CAAC,IAAI,CAAC,KAAK,CAAC,UAAU,CAAC,CAAC,CAAC;YACnD,SAAS;QACX,CAAC;QACD,eAAe,CAAC,KAAK,CAAC,CAAC;QACvB,IAAI,CAAC,KAAK,CAAC,CAAC;IACd,CAAC;AACH,CAAC;AAED;;;GAGG;AACH,MAAM,UAAU,YAAY,CAAC,IAAY;IACvC,MAAM,GAAG,GAAG,IAAI,SAAS,EAAE,CAAC,eAAe,CAAC,IAAI,EAAE,WAAW,CAAC,CAAC;IAC/D,IAAI,CAAC,GAAG,CAAC,IAAI,CAAC,CAAC;IACf,OAAO,GAAG,CAAC,IAAI,CAAC,SAAS,CAAC;AAC5B,CAAC","sourcesContent":["/**\n * HTML sanitization pre-pass. The ProseMirror schema is the real whitelist —\n * anything it can't parse is dropped — but this pass removes actively\n * dangerous content (scripts, event handlers, javascript: URLs) before the\n * string ever touches a live DOM, and strips style declarations down to the\n * set the schema understands.\n */\n\nimport { isAllowedEmbedSrc } from './plugins/embeds';\n\n/** Elements removed together with their content. */\nconst DROP_WITH_CONTENT = new Set([\n  'script', 'style', 'frame', 'frameset', 'object', 'embed',\n  'applet', 'form', 'input', 'textarea', 'select', 'button', 'meta', 'link',\n  'base', 'title', 'noscript', 'template', 'svg', 'math', 'canvas', 'audio',\n  'video', 'dialog', 'slot',\n]);\n\n/** Elements kept as-is (with filtered attributes). Extended in later phases. */\nconst ALLOWED_TAGS = new Set([\n  'p', 'h1', 'h2', 'h3', 'h4', 'h5', 'h6', 'blockquote', 'ul', 'ol', 'li',\n  'br', 'hr', 'a', 'b', 'strong', 'i', 'em', 'u', 's', 'strike', 'del',\n  'span', 'code', 'sub', 'sup', 'div', 'font',\n  'pre', 'figure', 'figcaption', 'img', 'iframe',\n  'table', 'thead', 'tbody', 'tfoot', 'tr', 'td', 'th', 'colgroup', 'col',\n]);\n\n/** Attributes allowed per tag (on top of GLOBAL_ATTRS). */\nconst TAG_ATTRS: Record<string, Set<string>> = {\n  a: new Set(['href', 'title', 'target', 'rel', 'id', 'name', 'data-bookmark']),\n  table: new Set(['role', 'data-layout']),\n  ol: new Set(['start', 'type', 'data-list-style']),\n  ul: new Set(['data-task-list']),\n  figure: new Set(['data-align']),\n  li: new Set(['data-task-item', 'data-checked']),\n  span: new Set([\n    'data-mention-trigger', 'data-mention-key', 'data-mention-label',\n    'data-merge-field', 'data-merge-label', 'data-math', 'data-math-display',\n  ]),\n  pre: new Set(['data-language']),\n  img: new Set(['src', 'alt', 'title', 'width', 'height']),\n  td: new Set(['colspan', 'rowspan', 'data-colwidth']),\n  th: new Set(['colspan', 'rowspan', 'data-colwidth']),\n  col: new Set(['span', 'width']),\n  div: new Set(['data-embed-provider', 'data-width', 'data-align']),\n  iframe: new Set(['src', 'title', 'frameborder', 'allowfullscreen', 'allow']),\n};\n\nconst GLOBAL_ATTRS = new Set(['style', 'align']);\n\n/** Style declarations the schema can represent. Everything else is noise. */\nconst ALLOWED_STYLES = new Set([\n  'color', 'background-color', 'font-family', 'font-size', 'font-style',\n  'font-weight', 'text-align', 'text-decoration', 'text-decoration-line',\n  'vertical-align', 'margin-left', 'margin-inline-start', 'list-style-type',\n  'line-height',\n]);\n\nconst SAFE_URL_PROTOCOLS = ['http:', 'https:', 'mailto:', 'tel:', 'ftp:'];\n\nexport function isSafeUrl(url: string, allowDataImage = false): boolean {\n  const trimmed = url.trim();\n  if (trimmed === '') return false;\n  // Relative URLs and fragments have no protocol — safe.\n  if (/^(#|\\/|\\.\\/|\\.\\.\\/)/.test(trimmed)) return true;\n  const match = /^([a-z][a-z0-9+.-]*):/i.exec(trimmed);\n  if (!match) return true;\n  const protocol = `${match[1].toLowerCase()}:`;\n  if (SAFE_URL_PROTOCOLS.includes(protocol)) return true;\n  if (allowDataImage && protocol === 'data:') {\n    // Raster images only — data:image/svg+xml can carry scripts.\n    return /^data:image\\/(png|gif|jpe?g|webp|avif|bmp);/i.test(trimmed);\n  }\n  return false;\n}\n\n/** Size styles allowed only where they are part of the document model. */\nconst SIZED_TAGS: Record<string, Set<string>> = {\n  img: new Set(['width', 'height']),\n  figure: new Set(['width', 'margin-right']),\n  div: new Set(['width', 'max-width']),\n};\n\nfunction filterStyle(value: string, tag: string): string {\n  const extra = SIZED_TAGS[tag];\n  return value\n    .split(';')\n    .map(decl => decl.trim())\n    .filter(decl => {\n      const name = decl.slice(0, decl.indexOf(':')).trim().toLowerCase();\n      if (!ALLOWED_STYLES.has(name) && !(extra && extra.has(name))) return false;\n      // Defense in depth: CSS url()/expression() smuggling.\n      return !/url\\s*\\(|expression\\s*\\(/i.test(decl);\n    })\n    .join('; ');\n}\n\nfunction sanitizeElement(el: Element): void {\n  const tag = el.tagName.toLowerCase();\n  const allowed = TAG_ATTRS[tag];\n  for (const attr of Array.from(el.attributes)) {\n    const name = attr.name.toLowerCase();\n    const keep =\n      (GLOBAL_ATTRS.has(name) || (allowed !== undefined && allowed.has(name))) &&\n      !name.startsWith('on');\n    if (!keep) {\n      el.removeAttribute(attr.name);\n      continue;\n    }\n    if (name === 'style') {\n      const filtered = filterStyle(attr.value, tag);\n      if (filtered) el.setAttribute('style', filtered);\n      else el.removeAttribute('style');\n    } else if (name === 'href' || name === 'src') {\n      // Raster data: URLs are allowed on images only.\n      if (!isSafeUrl(attr.value, tag === 'img')) el.removeAttribute(attr.name);\n    } else if (name === 'target') {\n      if (attr.value === '_blank') el.setAttribute('rel', 'noopener noreferrer');\n      else el.removeAttribute('target');\n    }\n  }\n}\n\nfunction walk(node: Element): void {\n  for (const child of Array.from(node.children)) {\n    const tag = child.tagName.toLowerCase();\n    if (DROP_WITH_CONTENT.has(tag)) {\n      child.remove();\n      continue;\n    }\n    // iframes are allowed ONLY for allowlisted embed providers.\n    if (tag === 'iframe' && !isAllowedEmbedSrc(child.getAttribute('src') ?? '')) {\n      child.remove();\n      continue;\n    }\n    if (!ALLOWED_TAGS.has(tag)) {\n      // Unknown wrapper (article, section, table for now, …): keep its\n      // content, drop the element itself.\n      walk(child);\n      child.replaceWith(...Array.from(child.childNodes));\n      continue;\n    }\n    sanitizeElement(child);\n    walk(child);\n  }\n}\n\n/**\n * Sanitizes an HTML string. Parsing happens in an inert document created by\n * DOMParser, so embedded scripts never execute and images never load.\n */\nexport function sanitizeHtml(html: string): string {\n  const doc = new DOMParser().parseFromString(html, 'text/html');\n  walk(doc.body);\n  return doc.body.innerHTML;\n}\n"]}
+{"version":3,"file":"sanitize.js","sourceRoot":"","sources":["../../../../src/nile-wysiwyg-editor/engine/sanitize.ts"],"names":[],"mappings":"AAAA;;;;;;GAMG;AAEH,OAAO,EAAE,iBAAiB,EAAE,aAAa,EAAE,qBAAqB,EAAE,MAAM,kBAAkB,CAAC;AAE3F,oDAAoD;AACpD,MAAM,iBAAiB,GAAG,IAAI,GAAG,CAAC;IAChC,QAAQ,EAAE,OAAO,EAAE,UAAU,EAAE,QAAQ,EAAE,OAAO;IAChD,QAAQ,EAAE,MAAM,EAAE,OAAO,EAAE,UAAU,EAAE,QAAQ,EAAE,QAAQ,EAAE,MAAM,EAAE,MAAM;IACzE,MAAM,EAAE,OAAO,EAAE,UAAU,EAAE,UAAU,EAAE,KAAK,EAAE,MAAM,EAAE,QAAQ,EAAE,OAAO;IACzE,OAAO,EAAE,QAAQ,EAAE,MAAM;CAC1B,CAAC,CAAC;AAEH,gFAAgF;AAChF,MAAM,YAAY,GAAG,IAAI,GAAG,CAAC;IAC3B,GAAG,EAAE,IAAI,EAAE,IAAI,EAAE,IAAI,EAAE,IAAI,EAAE,IAAI,EAAE,IAAI,EAAE,YAAY,EAAE,IAAI,EAAE,IAAI,EAAE,IAAI;IACvE,IAAI,EAAE,IAAI,EAAE,GAAG,EAAE,GAAG,EAAE,QAAQ,EAAE,GAAG,EAAE,IAAI,EAAE,GAAG,EAAE,GAAG,EAAE,QAAQ,EAAE,KAAK;IACpE,MAAM,EAAE,MAAM,EAAE,KAAK,EAAE,KAAK,EAAE,KAAK,EAAE,MAAM;IAC3C,KAAK,EAAE,QAAQ,EAAE,YAAY,EAAE,KAAK,EAAE,QAAQ;IAC9C,OAAO,EAAE,OAAO,EAAE,OAAO,EAAE,OAAO,EAAE,IAAI,EAAE,IAAI,EAAE,IAAI,EAAE,UAAU,EAAE,KAAK;CACxE,CAAC,CAAC;AAEH,2DAA2D;AAC3D,MAAM,SAAS,GAAgC;IAC7C,CAAC,EAAE,IAAI,GAAG,CAAC,CAAC,MAAM,EAAE,OAAO,EAAE,QAAQ,EAAE,KAAK,EAAE,IAAI,EAAE,MAAM,EAAE,eAAe,CAAC,CAAC;IAC7E,KAAK,EAAE,IAAI,GAAG,CAAC,CAAC,MAAM,EAAE,aAAa,CAAC,CAAC;IACvC,EAAE,EAAE,IAAI,GAAG,CAAC,CAAC,OAAO,EAAE,MAAM,EAAE,iBAAiB,CAAC,CAAC;IACjD,EAAE,EAAE,IAAI,GAAG,CAAC,CAAC,gBAAgB,CAAC,CAAC;IAC/B,MAAM,EAAE,IAAI,GAAG,CAAC,CAAC,YAAY,CAAC,CAAC;IAC/B,EAAE,EAAE,IAAI,GAAG,CAAC,CAAC,gBAAgB,EAAE,cAAc,CAAC,CAAC;IAC/C,IAAI,EAAE,IAAI,GAAG,CAAC;QACZ,sBAAsB,EAAE,kBAAkB,EAAE,oBAAoB;QAChE,kBAAkB,EAAE,kBAAkB,EAAE,WAAW,EAAE,mBAAmB;KACzE,CAAC;IACF,GAAG,EAAE,IAAI,GAAG,CAAC,CAAC,eAAe,CAAC,CAAC;IAC/B,GAAG,EAAE,IAAI,GAAG,CAAC,CAAC,KAAK,EAAE,KAAK,EAAE,OAAO,EAAE,OAAO,EAAE,QAAQ,CAAC,CAAC;IACxD,EAAE,EAAE,IAAI,GAAG,CAAC,CAAC,SAAS,EAAE,SAAS,EAAE,eAAe,CAAC,CAAC;IACpD,EAAE,EAAE,IAAI,GAAG,CAAC,CAAC,SAAS,EAAE,SAAS,EAAE,eAAe,CAAC,CAAC;IACpD,GAAG,EAAE,IAAI,GAAG,CAAC,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;IAC/B,GAAG,EAAE,IAAI,GAAG,CAAC,CAAC,qBAAqB,EAAE,YAAY,EAAE,YAAY,CAAC,CAAC;IACjE,MAAM,EAAE,IAAI,GAAG,CAAC;QACd,KAAK,EAAE,OAAO,EAAE,aAAa,EAAE,iBAAiB,EAAE,OAAO,EAAE,SAAS,EAAE,gBAAgB;KACvF,CAAC;CACH,CAAC;AAEF,MAAM,YAAY,GAAG,IAAI,GAAG,CAAC,CAAC,OAAO,EAAE,OAAO,EAAE,OAAO,EAAE,IAAI,CAAC,CAAC,CAAC;AAEhE;sDACsD;AACtD,MAAM,aAAa,GAAG,+BAA+B,CAAC;AAEtD,kFAAkF;AAClF,SAAS,cAAc,CAAC,KAAa;IACnC,OAAO,KAAK;SACT,KAAK,CAAC,KAAK,CAAC;SACZ,MAAM,CAAC,KAAK,CAAC,EAAE,CAAC,KAAK,IAAI,CAAC,aAAa,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC;SACpD,IAAI,CAAC,GAAG,CAAC,CAAC;AACf,CAAC;AAED,6EAA6E;AAC7E,MAAM,cAAc,GAAG,IAAI,GAAG,CAAC;IAC7B,OAAO,EAAE,kBAAkB,EAAE,aAAa,EAAE,WAAW,EAAE,YAAY;IACrE,aAAa,EAAE,YAAY,EAAE,iBAAiB,EAAE,sBAAsB;IACtE,gBAAgB,EAAE,aAAa,EAAE,qBAAqB,EAAE,iBAAiB;IACzE,aAAa;CACd,CAAC,CAAC;AAEH,MAAM,kBAAkB,GAAG,CAAC,OAAO,EAAE,QAAQ,EAAE,SAAS,EAAE,MAAM,EAAE,MAAM,CAAC,CAAC;AAE1E,MAAM,UAAU,SAAS,CAAC,GAAW,EAAE,cAAc,GAAG,KAAK;IAC3D,MAAM,OAAO,GAAG,GAAG,CAAC,IAAI,EAAE,CAAC;IAC3B,IAAI,OAAO,KAAK,EAAE;QAAE,OAAO,KAAK,CAAC;IACjC,uDAAuD;IACvD,IAAI,qBAAqB,CAAC,IAAI,CAAC,OAAO,CAAC;QAAE,OAAO,IAAI,CAAC;IACrD,MAAM,KAAK,GAAG,wBAAwB,CAAC,IAAI,CAAC,OAAO,CAAC,CAAC;IACrD,IAAI,CAAC,KAAK;QAAE,OAAO,IAAI,CAAC;IACxB,MAAM,QAAQ,GAAG,GAAG,KAAK,CAAC,CAAC,CAAC,CAAC,WAAW,EAAE,GAAG,CAAC;IAC9C,IAAI,kBAAkB,CAAC,QAAQ,CAAC,QAAQ,CAAC;QAAE,OAAO,IAAI,CAAC;IACvD,IAAI,cAAc,IAAI,QAAQ,KAAK,OAAO,EAAE,CAAC;QAC3C,6DAA6D;QAC7D,OAAO,8CAA8C,CAAC,IAAI,CAAC,OAAO,CAAC,CAAC;IACtE,CAAC;IACD,OAAO,KAAK,CAAC;AACf,CAAC;AAED,0EAA0E;AAC1E,MAAM,UAAU,GAAgC;IAC9C,GAAG,EAAE,IAAI,GAAG,CAAC,CAAC,OAAO,EAAE,QAAQ,CAAC,CAAC;IACjC,MAAM,EAAE,IAAI,GAAG,CAAC,CAAC,OAAO,EAAE,cAAc,CAAC,CAAC;IAC1C,GAAG,EAAE,IAAI,GAAG,CAAC,CAAC,OAAO,EAAE,WAAW,CAAC,CAAC;CACrC,CAAC;AAEF,SAAS,WAAW,CAAC,KAAa,EAAE,GAAW;IAC7C,MAAM,KAAK,GAAG,UAAU,CAAC,GAAG,CAAC,CAAC;IAC9B,OAAO,KAAK;SACT,KAAK,CAAC,GAAG,CAAC;SACV,GAAG,CAAC,IAAI,CAAC,EAAE,CAAC,IAAI,CAAC,IAAI,EAAE,CAAC;SACxB,MAAM,CAAC,IAAI,CAAC,EAAE;QACb,MAAM,IAAI,GAAG,IAAI,CAAC,KAAK,CAAC,CAAC,EAAE,IAAI,CAAC,OAAO,CAAC,GAAG,CAAC,CAAC,CAAC,IAAI,EAAE,CAAC,WAAW,EAAE,CAAC;QACnE,IAAI,CAAC,cAAc,CAAC,GAAG,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC,KAAK,IAAI,KAAK,CAAC,GAAG,CAAC,IAAI,CAAC,CAAC;YAAE,OAAO,KAAK,CAAC;QAC3E,sDAAsD;QACtD,OAAO,CAAC,2BAA2B,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;IACjD,CAAC,CAAC;SACD,IAAI,CAAC,IAAI,CAAC,CAAC;AAChB,CAAC;AAED;;;;;;;;;;GAUG;AACH,MAAM,UAAU,WAAW,CAAC,GAAW;IACrC,OAAO,GAAG;SACP,OAAO,CAAC,YAAY,EAAE,EAAE,CAAC;SACzB,OAAO,CAAC,mBAAmB,EAAE,EAAE,CAAC;SAChC,OAAO,CAAC,kBAAkB,EAAE,EAAE,CAAC;SAC/B,OAAO,CAAC,mBAAmB,EAAE,UAAU,CAAC;SACxC,OAAO,CAAC,iCAAiC,EAAE,iBAAiB,CAAC;SAC7D,OAAO,CAAC,kCAAkC,EAAE,CAAC,IAAI,EAAE,MAAM,EAAE,GAAG,EAAE,EAAE;QACjE,MAAM,GAAG,GAAG,MAAM,CAAC,GAAG,CAAC,CAAC,IAAI,EAAE,CAAC;QAC/B,IAAI,IAAI,CAAC,IAAI,CAAC,GAAG,CAAC;YAAE,OAAO,IAAI,CAAC;QAChC,IAAI,8CAA8C,CAAC,IAAI,CAAC,GAAG,CAAC;YAAE,OAAO,IAAI,CAAC;QAC1E,OAAO,QAAQ,CAAC;IAClB,CAAC,CAAC;SACD,IAAI,EAAE,CAAC;AACZ,CAAC;AAED,SAAS,eAAe,CAAC,EAAW;IAClC,MAAM,GAAG,GAAG,EAAE,CAAC,OAAO,CAAC,WAAW,EAAE,CAAC;IACrC,MAAM,OAAO,GAAG,SAAS,CAAC,GAAG,CAAC,CAAC;IAC/B,KAAK,MAAM,IAAI,IAAI,KAAK,CAAC,IAAI,CAAC,EAAE,CAAC,UAAU,CAAC,EAAE,CAAC;QAC7C,MAAM,IAAI,GAAG,IAAI,CAAC,IAAI,CAAC,WAAW,EAAE,CAAC;QACrC,MAAM,IAAI,GACR,CAAC,YAAY,CAAC,GAAG,CAAC,IAAI,CAAC,IAAI,CAAC,OAAO,KAAK,SAAS,IAAI,OAAO,CAAC,GAAG,CAAC,IAAI,CAAC,CAAC,CAAC;YACxE,CAAC,IAAI,CAAC,UAAU,CAAC,IAAI,CAAC,CAAC;QACzB,IAAI,CAAC,IAAI,EAAE,CAAC;YACV,EAAE,CAAC,eAAe,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;YAC9B,SAAS;QACX,CAAC;QACD,IAAI,IAAI,KAAK,OAAO,EAAE,CAAC;YACrB,MAAM,QAAQ,GAAG,WAAW,CAAC,IAAI,CAAC,KAAK,EAAE,GAAG,CAAC,CAAC;YAC9C,IAAI,QAAQ;gBAAE,EAAE,CAAC,YAAY,CAAC,OAAO,EAAE,QAAQ,CAAC,CAAC;;gBAC5C,EAAE,CAAC,eAAe,CAAC,OAAO,CAAC,CAAC;QACnC,CAAC;aAAM,IAAI,IAAI,KAAK,OAAO,EAAE,CAAC;YAC5B,MAAM,OAAO,GAAG,cAAc,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC;YAC3C,IAAI,OAAO;gBAAE,EAAE,CAAC,YAAY,CAAC,OAAO,EAAE,OAAO,CAAC,CAAC;;gBAC1C,EAAE,CAAC,eAAe,CAAC,OAAO,CAAC,CAAC;QACnC,CAAC;aAAM,IAAI,IAAI,KAAK,IAAI,EAAE,CAAC;YACzB,IAAI,aAAa,CAAC,IAAI,CAAC,IAAI,CAAC,KAAK,CAAC;gBAAE,EAAE,CAAC,eAAe,CAAC,IAAI,CAAC,CAAC;QAC/D,CAAC;aAAM,IAAI,IAAI,KAAK,MAAM,IAAI,IAAI,KAAK,KAAK,EAAE,CAAC;YAC7C,gDAAgD;YAChD,IAAI,CAAC,SAAS,CAAC,IAAI,CAAC,KAAK,EAAE,GAAG,KAAK,KAAK,CAAC;gBAAE,EAAE,CAAC,eAAe,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;QAC3E,CAAC;aAAM,IAAI,IAAI,KAAK,QAAQ,EAAE,CAAC;YAC7B,IAAI,IAAI,CAAC,KAAK,KAAK,QAAQ;gBAAE,EAAE,CAAC,YAAY,CAAC,KAAK,EAAE,qBAAqB,CAAC,CAAC;;gBACtE,EAAE,CAAC,eAAe,CAAC,QAAQ,CAAC,CAAC;QACpC,CAAC;IACH,CAAC;AACH,CAAC;AAED,SAAS,IAAI,CAAC,IAAa;IACzB,KAAK,MAAM,KAAK,IAAI,KAAK,CAAC,IAAI,CAAC,IAAI,CAAC,QAAQ,CAAC,EAAE,CAAC;QAC9C,MAAM,GAAG,GAAG,KAAK,CAAC,OAAO,CAAC,WAAW,EAAE,CAAC;QACxC,IAAI,iBAAiB,CAAC,GAAG,CAAC,GAAG,CAAC,EAAE,CAAC;YAC/B,KAAK,CAAC,MAAM,EAAE,CAAC;YACf,SAAS;QACX,CAAC;QACD,0EAA0E;QAC1E,yEAAyE;QACzE,IAAI,GAAG,KAAK,QAAQ,EAAE,CAAC;YACrB,IAAI,CAAC,iBAAiB,CAAC,KAAK,CAAC,YAAY,CAAC,KAAK,CAAC,IAAI,EAAE,CAAC,EAAE,CAAC;gBACxD,KAAK,CAAC,MAAM,EAAE,CAAC;gBACf,SAAS;YACX,CAAC;YACD,KAAK,CAAC,YAAY,CAAC,SAAS,EAAE,aAAa,CAAC,CAAC;YAC7C,KAAK,CAAC,YAAY,CAAC,gBAAgB,EAAE,qBAAqB,CAAC,CAAC;QAC9D,CAAC;QACD,yEAAyE;QACzE,2EAA2E;QAC3E,8DAA8D;QAC9D,IAAI,GAAG,KAAK,OAAO,EAAE,CAAC;YACpB,KAAK,CAAC,WAAW,GAAG,WAAW,CAAC,KAAK,CAAC,WAAW,IAAI,EAAE,CAAC,CAAC;YACzD,KAAK,MAAM,IAAI,IAAI,KAAK,CAAC,IAAI,CAAC,KAAK,CAAC,UAAU,CAAC;gBAAE,KAAK,CAAC,eAAe,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;YAClF,SAAS;QACX,CAAC;QACD,IAAI,CAAC,YAAY,CAAC,GAAG,CAAC,GAAG,CAAC,EAAE,CAAC;YAC3B,iEAAiE;YACjE,oCAAoC;YACpC,IAAI,CAAC,KAAK,CAAC,CAAC;YACZ,KAAK,CAAC,WAAW,CAAC,GAAG,KAAK,CAAC,IAAI,CAAC,KAAK,CAAC,UAAU,CAAC,CAAC,CAAC;YACnD,SAAS;QACX,CAAC;QACD,eAAe,CAAC,KAAK,CAAC,CAAC;QACvB,IAAI,CAAC,KAAK,CAAC,CAAC;IACd,CAAC;AACH,CAAC;AAED;;;GAGG;AACH,MAAM,UAAU,YAAY,CAAC,IAAY;IACvC,6EAA6E;IAC7E,2EAA2E;IAC3E,2EAA2E;IAC3E,MAAM,GAAG,GAAG,IAAI,SAAS,EAAE,CAAC,eAAe,CAAC,SAAS,IAAI,SAAS,EAAE,WAAW,CAAC,CAAC;IACjF,IAAI,CAAC,GAAG,CAAC,IAAI,CAAC,CAAC;IACf,OAAO,GAAG,CAAC,IAAI,CAAC,SAAS,CAAC;AAC5B,CAAC","sourcesContent":["/**\n * HTML sanitization pre-pass. The ProseMirror schema is the real whitelist —\n * anything it can't parse is dropped — but this pass removes actively\n * dangerous content (scripts, event handlers, javascript: URLs) before the\n * string ever touches a live DOM, and strips style declarations down to the\n * set the schema understands.\n */\n\nimport { isAllowedEmbedSrc, EMBED_SANDBOX, EMBED_REFERRER_POLICY } from './plugins/embeds';\n\n/** Elements removed together with their content. */\nconst DROP_WITH_CONTENT = new Set([\n  'script', 'frame', 'frameset', 'object', 'embed',\n  'applet', 'form', 'input', 'textarea', 'select', 'button', 'meta', 'link',\n  'base', 'title', 'noscript', 'template', 'svg', 'math', 'canvas', 'audio',\n  'video', 'dialog', 'slot',\n]);\n\n/** Elements kept as-is (with filtered attributes). Extended in later phases. */\nconst ALLOWED_TAGS = new Set([\n  'p', 'h1', 'h2', 'h3', 'h4', 'h5', 'h6', 'blockquote', 'ul', 'ol', 'li',\n  'br', 'hr', 'a', 'b', 'strong', 'i', 'em', 'u', 's', 'strike', 'del',\n  'span', 'code', 'sub', 'sup', 'div', 'font',\n  'pre', 'figure', 'figcaption', 'img', 'iframe',\n  'table', 'thead', 'tbody', 'tfoot', 'tr', 'td', 'th', 'colgroup', 'col',\n]);\n\n/** Attributes allowed per tag (on top of GLOBAL_ATTRS). */\nconst TAG_ATTRS: Record<string, Set<string>> = {\n  a: new Set(['href', 'title', 'target', 'rel', 'id', 'name', 'data-bookmark']),\n  table: new Set(['role', 'data-layout']),\n  ol: new Set(['start', 'type', 'data-list-style']),\n  ul: new Set(['data-task-list']),\n  figure: new Set(['data-align']),\n  li: new Set(['data-task-item', 'data-checked']),\n  span: new Set([\n    'data-mention-trigger', 'data-mention-key', 'data-mention-label',\n    'data-merge-field', 'data-merge-label', 'data-math', 'data-math-display',\n  ]),\n  pre: new Set(['data-language']),\n  img: new Set(['src', 'alt', 'title', 'width', 'height']),\n  td: new Set(['colspan', 'rowspan', 'data-colwidth']),\n  th: new Set(['colspan', 'rowspan', 'data-colwidth']),\n  col: new Set(['span', 'width']),\n  div: new Set(['data-embed-provider', 'data-width', 'data-align']),\n  iframe: new Set([\n    'src', 'title', 'frameborder', 'allowfullscreen', 'allow', 'sandbox', 'referrerpolicy',\n  ]),\n};\n\nconst GLOBAL_ATTRS = new Set(['style', 'align', 'class', 'id']);\n\n/** Author class/id names starting with these are the editor's own; drop them so\n * content can't impersonate or restyle the chrome. */\nconst RESERVED_NAME = /^(?:nile-wysiwyg|ProseMirror)/;\n\n/** Filtered `class`: reserved-prefixed tokens removed. Empty → drop attribute. */\nfunction cleanClassAttr(value: string): string {\n  return value\n    .split(/\\s+/)\n    .filter(token => token && !RESERVED_NAME.test(token))\n    .join(' ');\n}\n\n/** Style declarations the schema can represent. Everything else is noise. */\nconst ALLOWED_STYLES = new Set([\n  'color', 'background-color', 'font-family', 'font-size', 'font-style',\n  'font-weight', 'text-align', 'text-decoration', 'text-decoration-line',\n  'vertical-align', 'margin-left', 'margin-inline-start', 'list-style-type',\n  'line-height',\n]);\n\nconst SAFE_URL_PROTOCOLS = ['http:', 'https:', 'mailto:', 'tel:', 'ftp:'];\n\nexport function isSafeUrl(url: string, allowDataImage = false): boolean {\n  const trimmed = url.trim();\n  if (trimmed === '') return false;\n  // Relative URLs and fragments have no protocol — safe.\n  if (/^(#|\\/|\\.\\/|\\.\\.\\/)/.test(trimmed)) return true;\n  const match = /^([a-z][a-z0-9+.-]*):/i.exec(trimmed);\n  if (!match) return true;\n  const protocol = `${match[1].toLowerCase()}:`;\n  if (SAFE_URL_PROTOCOLS.includes(protocol)) return true;\n  if (allowDataImage && protocol === 'data:') {\n    // Raster images only — data:image/svg+xml can carry scripts.\n    return /^data:image\\/(png|gif|jpe?g|webp|avif|bmp);/i.test(trimmed);\n  }\n  return false;\n}\n\n/** Size styles allowed only where they are part of the document model. */\nconst SIZED_TAGS: Record<string, Set<string>> = {\n  img: new Set(['width', 'height']),\n  figure: new Set(['width', 'margin-right']),\n  div: new Set(['width', 'max-width']),\n};\n\nfunction filterStyle(value: string, tag: string): string {\n  const extra = SIZED_TAGS[tag];\n  return value\n    .split(';')\n    .map(decl => decl.trim())\n    .filter(decl => {\n      const name = decl.slice(0, decl.indexOf(':')).trim().toLowerCase();\n      if (!ALLOWED_STYLES.has(name) && !(extra && extra.has(name))) return false;\n      // Defense in depth: CSS url()/expression() smuggling.\n      return !/url\\s*\\(|expression\\s*\\(/i.test(decl);\n    })\n    .join('; ');\n}\n\n/**\n * Sanitizes the CSS inside a <style> block. Neutralizes the vectors that make\n * a raw stylesheet dangerous while leaving ordinary rules intact:\n *  - `@import` (external fetch / data exfiltration) — removed,\n *  - `expression()`, `behavior:`, `-moz-binding:` (script execution) — broken,\n *  - `url(...)` pointing anywhere other than a safe data: image or a local\n *    `#fragment` — replaced with `url(#)`,\n *  - any `</style` sequence — stripped so the text can't break out of the\n *    element when the document is re-serialized.\n * CSS comments are dropped first so none of the above can hide inside one.\n */\nexport function sanitizeCss(css: string): string {\n  return css\n    .replace(/<\\/style/gi, '')\n    .replace(/\\/\\*[\\s\\S]*?\\*\\//g, '')\n    .replace(/@import[^;]*;?/gi, '')\n    .replace(/expression\\s*\\(/gi, 'blocked(')\n    .replace(/(behavior|-moz-binding)(\\s*):/gi, 'blocked-prop$2:')\n    .replace(/url\\(\\s*(['\"]?)([^)]*?)\\1\\s*\\)/gi, (full, _quote, raw) => {\n      const url = String(raw).trim();\n      if (/^#/.test(url)) return full;\n      if (/^data:image\\/(png|gif|jpe?g|webp|avif|bmp);/i.test(url)) return full;\n      return 'url(#)';\n    })\n    .trim();\n}\n\nfunction sanitizeElement(el: Element): void {\n  const tag = el.tagName.toLowerCase();\n  const allowed = TAG_ATTRS[tag];\n  for (const attr of Array.from(el.attributes)) {\n    const name = attr.name.toLowerCase();\n    const keep =\n      (GLOBAL_ATTRS.has(name) || (allowed !== undefined && allowed.has(name))) &&\n      !name.startsWith('on');\n    if (!keep) {\n      el.removeAttribute(attr.name);\n      continue;\n    }\n    if (name === 'style') {\n      const filtered = filterStyle(attr.value, tag);\n      if (filtered) el.setAttribute('style', filtered);\n      else el.removeAttribute('style');\n    } else if (name === 'class') {\n      const cleaned = cleanClassAttr(attr.value);\n      if (cleaned) el.setAttribute('class', cleaned);\n      else el.removeAttribute('class');\n    } else if (name === 'id') {\n      if (RESERVED_NAME.test(attr.value)) el.removeAttribute('id');\n    } else if (name === 'href' || name === 'src') {\n      // Raster data: URLs are allowed on images only.\n      if (!isSafeUrl(attr.value, tag === 'img')) el.removeAttribute(attr.name);\n    } else if (name === 'target') {\n      if (attr.value === '_blank') el.setAttribute('rel', 'noopener noreferrer');\n      else el.removeAttribute('target');\n    }\n  }\n}\n\nfunction walk(node: Element): void {\n  for (const child of Array.from(node.children)) {\n    const tag = child.tagName.toLowerCase();\n    if (DROP_WITH_CONTENT.has(tag)) {\n      child.remove();\n      continue;\n    }\n    // iframes are allowed for any https src, but always force-sandboxed so an\n    // untrusted embed can never navigate the top window or run unrestricted.\n    if (tag === 'iframe') {\n      if (!isAllowedEmbedSrc(child.getAttribute('src') ?? '')) {\n        child.remove();\n        continue;\n      }\n      child.setAttribute('sandbox', EMBED_SANDBOX);\n      child.setAttribute('referrerpolicy', EMBED_REFERRER_POLICY);\n    }\n    // <style> blocks are kept, but their CSS is sanitized and any attributes\n    // (media, nonce, …) dropped. Handled before the unwrap branch below so the\n    // raw CSS is never spilled into the document as visible text.\n    if (tag === 'style') {\n      child.textContent = sanitizeCss(child.textContent ?? '');\n      for (const attr of Array.from(child.attributes)) child.removeAttribute(attr.name);\n      continue;\n    }\n    if (!ALLOWED_TAGS.has(tag)) {\n      // Unknown wrapper (article, section, table for now, …): keep its\n      // content, drop the element itself.\n      walk(child);\n      child.replaceWith(...Array.from(child.childNodes));\n      continue;\n    }\n    sanitizeElement(child);\n    walk(child);\n  }\n}\n\n/**\n * Sanitizes an HTML string. Parsing happens in an inert document created by\n * DOMParser, so embedded scripts never execute and images never load.\n */\nexport function sanitizeHtml(html: string): string {\n  // Wrap in an explicit <body> so metadata-like tags (notably <style>) stay in\n  // the body instead of being hoisted into <head> by the HTML parser — which\n  // would drop them before walk() ever sees them. Mirrors htmlToDoc's parse.\n  const doc = new DOMParser().parseFromString(`<body>${html}</body>`, 'text/html');\n  walk(doc.body);\n  return doc.body.innerHTML;\n}\n"]}

package/dist/src/nile-wysiwyg-editor/engine/schema.d.ts

@@ -3,6 +3,8 @@ export type Alignment = 'left' | 'center' | 'right' | 'justify';
 /** Indent depth in 40px steps, like CKEditor/Google Docs. */
 export declare const INDENT_STEP_PX = 40;
 export declare const MAX_INDENT = 8;
+export declare function cleanClass(value: string | null): string | null;
+export declare function cleanId(value: string | null): string | null;
 /** Bullet markers CSS understands; anything else falls back to default. */
 export declare const BULLET_STYLES: readonly ["disc", "circle", "square"];
 /** Numbering styles, keyed by the HTML `type` attribute letter. */

package/dist/src/nile-wysiwyg-editor/engine/schema.js

@@ -2,6 +2,7 @@ import { Schema } from 'prosemirror-model';
 import { addListNodes } from 'prosemirror-schema-list';
 import { tableNodes } from 'prosemirror-tables';
 import OrderedMap from 'orderedmap';
+import { EMBED_SANDBOX, EMBED_REFERRER_POLICY } from './plugins/embeds';
 const ALIGN_VALUES = ['left', 'center', 'right', 'justify'];
 function getAlignFromDom(dom) {
     const value = dom.style.textAlign || dom.getAttribute('align') || '';
@@ -47,6 +48,95 @@ function blockStyle(attrs) {
         parts.push(`line-height: ${attrs.lineHeight}`);
     return parts.length ? parts.join('; ') : null;
 }
+/* -------------------------------------------------------------------------
+ * Author class / id passthrough
+ *
+ * Content elements keep the `class` and `id` an author gave them, so class/id
+ * CSS (e.g. a scoped author <style> block) can target the document. Values that
+ * collide with the editor's own internal names are dropped so content can never
+ * impersonate or restyle the chrome; the editor's own classes (emitted in
+ * toDOM) are always preserved and merged with the author's.
+ * ------------------------------------------------------------------------- */
+const RESERVED_NAME = /^(?:nile-wysiwyg|ProseMirror)/;
+export function cleanClass(value) {
+    if (!value)
+        return null;
+    const kept = value.split(/\s+/).filter(token => token && !RESERVED_NAME.test(token));
+    return kept.length ? kept.join(' ') : null;
+}
+export function cleanId(value) {
+    return value && !RESERVED_NAME.test(value) ? value : null;
+}
+const HTML_ATTR_SPEC = { htmlClass: { default: null }, htmlId: { default: null } };
+function readHtmlAttrs(dom) {
+    return {
+        htmlClass: cleanClass(dom.getAttribute('class')),
+        htmlId: cleanId(dom.getAttribute('id')),
+    };
+}
+/** Wraps a tag-based parse rule so it also captures author class/id. Rules
+ * keyed by `style` (whose getAttrs receives a value string, not an element)
+ * pass through untouched. */
+function augmentParseRule(rule) {
+    if (!('tag' in rule) || 'style' in rule)
+        return rule;
+    const inner = rule.getAttrs;
+    return {
+        ...rule,
+        getAttrs(dom) {
+            const base = inner ? inner(dom) : null;
+            if (base === false)
+                return false;
+            return { ...(base ?? {}), ...readHtmlAttrs(dom) };
+        },
+    };
+}
+/** Wraps a toDOM so stored class/id are written back onto the element, merged
+ * with any class the spec emits itself. Handles the `[tag]`, `[tag, attrs]`,
+ * `[tag, 0]`, and `[tag, attrs, …children]` output shapes. */
+function augmentToDOM(toDOM) {
+    if (!toDOM)
+        return toDOM;
+    return (node, ...rest) => {
+        const out = toDOM(node, ...rest);
+        const cls = node.attrs.htmlClass;
+        const id = node.attrs.htmlId;
+        if (!Array.isArray(out) || (!cls && !id))
+            return out;
+        const [tag, ...tail] = out;
+        let attrs = {};
+        let children = tail;
+        if (tail.length && tail[0] && typeof tail[0] === 'object' && !Array.isArray(tail[0])) {
+            attrs = { ...tail[0] };
+            children = tail.slice(1);
+        }
+        if (cls)
+            attrs.class = attrs.class ? `${attrs.class} ${cls}` : cls;
+        if (id)
+            attrs.id = id;
+        return [tag, attrs, ...children];
+    };
+}
+/** Adds `htmlClass`/`htmlId` attrs plus parse/serialize support to a spec. */
+function withHtmlAttrs(spec) {
+    return {
+        ...spec,
+        attrs: { ...(spec.attrs ?? {}), ...HTML_ATTR_SPEC },
+        parseDOM: spec.parseDOM?.map(augmentParseRule),
+        toDOM: augmentToDOM(spec.toDOM),
+    };
+}
+/** Content nodes/marks that carry author class/id. Excludes structural leaves
+ * (doc, text), atoms with bespoke DOM (image, embeds, mentions, math, …) and
+ * style-only marks (their spans are covered by `html_span`). */
+const HTML_ATTR_NODES = [
+    'paragraph', 'heading', 'blockquote', 'code_block',
+    'bullet_list', 'ordered_list', 'list_item',
+    'table', 'table_row', 'table_cell', 'table_header',
+];
+const HTML_ATTR_MARKS = [
+    'link', 'bold', 'italic', 'underline', 'strike', 'code', 'subscript', 'superscript',
+];
 /**
  * Block nodes. `align` is modeled as an attribute on textblocks rather than a
  * wrapper node so alignment survives block-type changes (heading <-> paragraph).
@@ -360,6 +450,8 @@ const nodes = {
                 frameborder: '0',
                 allowfullscreen: 'true',
                 allow: 'accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture',
+                sandbox: EMBED_SANDBOX,
+                referrerpolicy: EMBED_REFERRER_POLICY,
             };
             if (title)
                 iframeAttrs.title = title;
@@ -384,6 +476,28 @@ const nodes = {
             return ['div', wrapperAttrs, ['iframe', iframeAttrs]];
         },
     },
+    /**
+     * Author-supplied `<style>` block, entered via the HTML source view. The CSS
+     * is sanitized on the way in (see sanitizeCss) — `@import`, `expression()`
+     * and external `url()` are neutralized — then stored verbatim and serialized
+     * straight back to a `<style>` element. A node view renders it live in the
+     * editor with a small chip so the otherwise-invisible block stays selectable.
+     */
+    styleBlock: {
+        group: 'block',
+        atom: true,
+        selectable: true,
+        attrs: { css: { default: '' } },
+        parseDOM: [
+            {
+                tag: 'style',
+                getAttrs: (dom) => ({ css: dom.textContent ?? '' }),
+            },
+        ],
+        toDOM(node) {
+            return ['style', node.attrs.css];
+        },
+    },
     /**
      * Bookmark: a named in-text anchor (`<a id="…">`) for intra-document
      * navigation. Invisible in serialized output; the editor draws a flag
@@ -685,6 +799,31 @@ const marks = {
             return ['code', { class: 'nile-wysiwyg__inline-code' }, 0];
         },
     },
+    /** Carries an author `class`/`id` on an inline `<span>` that has no other
+     * meaning, so class/id CSS can target runs of text. Low priority so styled
+     * spans (color/font marks) and atom spans (mention/merge/math) win. */
+    html_span: {
+        attrs: { htmlClass: { default: null }, htmlId: { default: null } },
+        parseDOM: [
+            {
+                tag: 'span',
+                priority: 5,
+                getAttrs: (dom) => {
+                    const htmlClass = cleanClass(dom.getAttribute('class'));
+                    const htmlId = cleanId(dom.getAttribute('id'));
+                    return htmlClass || htmlId ? { htmlClass, htmlId } : false;
+                },
+            },
+        ],
+        toDOM(mark) {
+            const attrs = {};
+            if (mark.attrs.htmlClass)
+                attrs.class = mark.attrs.htmlClass;
+            if (mark.attrs.htmlId)
+                attrs.id = mark.attrs.htmlId;
+            return ['span', attrs, 0];
+        },
+    },
     text_color: {
         attrs: { color: {} },
         parseDOM: [
@@ -811,12 +950,25 @@ const orderedListSpec = {
 };
 /** Builds the editor schema. List nodes are added via prosemirror-schema-list. */
 export function createSchema() {
-    const nodeMap = addListNodes(OrderedMap.from(nodes), 'paragraph block*', 'block')
+    let nodeMap = addListNodes(OrderedMap.from(nodes), 'paragraph block*', 'block')
         .update('bullet_list', bulletListSpec)
         .update('ordered_list', orderedListSpec)
         .append(tableNodeSpecs)
         .update('table', tableSpec);
-    return new Schema({ nodes: nodeMap, marks: OrderedMap.from(marks) });
+    // Thread author class/id through the content nodes and marks (applied last so
+    // it also wraps the list/table specs pulled in from external packages).
+    for (const name of HTML_ATTR_NODES) {
+        const spec = nodeMap.get(name);
+        if (spec)
+            nodeMap = nodeMap.update(name, withHtmlAttrs(spec));
+    }
+    let markMap = OrderedMap.from(marks);
+    for (const name of HTML_ATTR_MARKS) {
+        const spec = markMap.get(name);
+        if (spec)
+            markMap = markMap.update(name, withHtmlAttrs(spec));
+    }
+    return new Schema({ nodes: nodeMap, marks: markMap });
 }
 /** A shared default schema instance. */
 export const defaultSchema = createSchema();