fix: replace control character sentinel with plain ASCII to prevent rendering artifacts

Author: chizkidd

docs/index.html

@@ -1950,10 +1950,11 @@ <h3 id="panel-heading">Contents</h3>
   }
   if (!container) container = doc.body;
 
+
   // Inject sentinel strings at <br> tags and block-level boundaries before
   // reading textContent. This handles sites that use <br><br> for paragraphs
   // (e.g. paulgraham.com) as well as sites that use <p> or <div> tags.
-  const SENTINEL = '\u0002\n\n';
+  const SENTINEL = '||INKCAST_BREAK||';
   const BLOCK_TAGS = new Set(['P','H1','H2','H3','H4','H5','H6','LI','BLOCKQUOTE','PRE','TD','TH','DIV','TR','SECTION','ARTICLE']);
   const tw = doc.createTreeWalker(container, NodeFilter.SHOW_ELEMENT);
   const toMark = [];
@@ -1966,7 +1967,8 @@ <h3 id="panel-heading">Contents</h3>
     .split(SENTINEL)
     .map(s => s.trim())
     .filter(Boolean)
-    .join('\n\n');
+    .join('\n\n')
+    .replace(/\|\|INKCAST_BREAK\|\|/g, '\n\n'); // strip any that leaked through
   text = text.replace(/\r\n/g, '\n').replace(/\r/g, '\n').replace(/\n{3,}/g, '\n\n').trim();
   if (text.length < 100) return null;