Trim links and include images

.gitignore

@@ -12,6 +12,9 @@ Sources/ContentScopeScripts/dist/
 test-results
 !Sources/ContentScopeScripts/dist/pages/.gitignore
 
+# Test output files (generated during tests)
+injected/unit-test/fixtures/page-context/output/
+
 # Local Netlify folder
 .netlify
 # VS Code user config

injected/package.json

@@ -26,12 +26,12 @@
   },
   "type": "module",
   "dependencies": {
+    "@duckduckgo/privacy-configuration": "github:duckduckgo/privacy-configuration#1752154773643",
+    "esbuild": "^0.25.10",
     "minimist": "^1.2.8",
     "parse-address": "^1.1.2",
     "seedrandom": "^3.0.5",
     "sjcl": "^1.0.8",
-    "@duckduckgo/privacy-configuration": "github:duckduckgo/privacy-configuration#1752154773643",
-    "esbuild": "^0.25.10",
     "urlpattern-polyfill": "^10.1.0"
   },
   "devDependencies": {
@@ -43,6 +43,7 @@
     "@typescript-eslint/eslint-plugin": "^8.46.0",
     "fast-check": "^4.2.0",
     "jasmine": "^5.12.0",
+    "jsdom": "^27.0.0",
     "web-ext": "^9.0.0"
   }
 }

injected/src/features/page-context.js

@@ -3,7 +3,7 @@ import { getFaviconList } from './favicon.js';
 import { isDuckAi, isBeingFramed, getTabUrl } from '../utils.js';
 const MSG_PAGE_CONTEXT_RESPONSE = 'collectionResult';
 
-function checkNodeIsVisible(node) {
+export function checkNodeIsVisible(node) {
     try {
         const style = window.getComputedStyle(node);
 
@@ -36,6 +36,29 @@ function isHtmlElement(node) {
  * @returns {Document | null}
  */
 function getSameOriginIframeDocument(iframe) {
+    // Pre-check conditions that would prevent access without triggering security errors
+    const src = iframe.src;
+
+    // Skip sandboxed iframes unless they explicitly allow scripts
+    // Avoids: Blocked script execution in 'about:blank' because the document's frame is sandboxed and the 'allow-scripts' permission is not set.
+    // Note: iframe.sandbox always returns a DOMTokenList, so check hasAttribute instead
+    if (iframe.hasAttribute('sandbox') && !iframe.sandbox.contains('allow-scripts')) {
+        return null;
+    }
+
+    // Check for cross-origin URLs (but allow about:blank and empty src as they inherit parent origin)
+    if (src && src !== 'about:blank' && src !== '') {
+        try {
+            const iframeUrl = new URL(src, window.location.href);
+            if (iframeUrl.origin !== window.location.origin) {
+                return null;
+            }
+        } catch (e) {
+            // Invalid URL, skip
+            return null;
+        }
+    }
+
     try {
         // Try to access the contentDocument - this will throw if cross-origin
         const doc = iframe.contentDocument;
@@ -76,8 +99,9 @@ function domToMarkdownChildren(childNodes, settings, depth = 0) {
  * @typedef {Object} DomToMarkdownSettings
  * @property {number} maxLength - Maximum length of content
  * @property {number} maxDepth - Maximum depth to traverse
- * @property {string} excludeSelectors - CSS selectors to exclude from processing
+ * @property {string | null} excludeSelectors - CSS selectors to exclude from processing
  * @property {boolean} includeIframes - Whether to include iframe content
+ * @property {boolean} trimBlankLinks - Whether to trim blank links
  */
 
 /**
@@ -87,7 +111,7 @@ function domToMarkdownChildren(childNodes, settings, depth = 0) {
  * @param {number} depth
  * @returns {string}
  */
-function domToMarkdown(node, settings, depth = 0) {
+export function domToMarkdown(node, settings, depth = 0) {
     if (depth > settings.maxDepth) {
         return '';
     }
@@ -97,7 +121,7 @@ function domToMarkdown(node, settings, depth = 0) {
     if (!isHtmlElement(node)) {
         return '';
     }
-    if (!checkNodeIsVisible(node) || node.matches(settings.excludeSelectors)) {
+    if (!checkNodeIsVisible(node) || (settings.excludeSelectors && node.matches(settings.excludeSelectors))) {
         return '';
     }
 
@@ -127,12 +151,15 @@ function domToMarkdown(node, settings, depth = 0) {
             return `${children}\n`;
         case 'br':
             return `\n`;
+        case 'img':
+            return `\n![${getAttributeOrBlank(node, 'alt')}](${getAttributeOrBlank(node, 'src')})\n`;
         case 'ul':
+        case 'ol':
             return `\n${children}\n`;
         case 'li':
-            return `\n- ${children.trim()}\n`;
+            return `\n- ${collapseAndTrim(children)}\n`;
         case 'a':
-            return getLinkText(node);
+            return getLinkText(node, children, settings);
         case 'iframe': {
             if (!settings.includeIframes) {
                 return children;
@@ -151,13 +178,30 @@ function domToMarkdown(node, settings, depth = 0) {
     }
 }
 
+/**
+ * @param {Element} node
+ * @param {string} attr
+ * @returns {string}
+ */
+function getAttributeOrBlank(node, attr) {
+    const attrValue = node.getAttribute(attr) ?? '';
+    return attrValue.trim();
+}
+
 function collapseAndTrim(str) {
     return collapseWhitespace(str).trim();
 }
 
-function getLinkText(node) {
+function getLinkText(node, children, settings) {
     const href = node.getAttribute('href');
-    return href ? `[${collapseAndTrim(node.textContent)}](${href})` : collapseWhitespace(node.textContent);
+    const trimmedContent = collapseAndTrim(children);
+    if (settings.trimBlankLinks && trimmedContent.length === 0) {
+        return '';
+    }
+    // The difference in whitespace handling is intentional here.
+    // Where we don't wrap in a link:
+    // we should retain at least one preceding and following space.
+    return href ? `[${trimmedContent}](${href})` : collapseWhitespace(children);
 }
 
 export default class PageContext extends ContentFeature {
@@ -420,6 +464,7 @@ export default class PageContext extends ContentFeature {
         const maxDepth = this.getFeatureSetting('maxDepth') || 5000;
         let excludeSelectors = this.getFeatureSetting('excludeSelectors') || ['.ad', '.sidebar', '.footer', '.nav', '.header'];
         const excludedInertElements = this.getFeatureSetting('excludedInertElements') || [
+            'img', // Note we're currently disabling images which we're handling in domToMarkdown (this can be per-site enabled in the config if needed).
             'script',
             'style',
             'link',
@@ -436,22 +481,34 @@ export default class PageContext extends ContentFeature {
         const mainContentSelector = this.getFeatureSetting('mainContentSelector') || 'main, article, .content, .main, #content, #main';
         let mainContent = document.querySelector(mainContentSelector);
         const mainContentLength = this.getFeatureSetting('mainContentLength') || 100;
+        // Fast path to avoid processing main content if it's too short
         if (mainContent && mainContent.innerHTML.trim().length <= mainContentLength) {
             mainContent = null;
         }
-        const contentRoot = mainContent || document.body;
+        let contentRoot = mainContent || document.body;
 
-        if (contentRoot) {
-            this.log.info('Getting main content', contentRoot);
-            content += domToMarkdown(contentRoot, {
+        // Use a closure to reuse the domToMarkdown parameters
+        const extractContent = (root) => {
+            this.log.info('Getting content', root);
+            const result = domToMarkdown(root, {
                 maxLength: upperLimit,
                 maxDepth,
                 includeIframes: this.getFeatureSettingEnabled('includeIframes', 'enabled'),
                 excludeSelectors: excludeSelectorsString,
-            });
-            this.log.info('Content markdown', content, contentRoot);
+                trimBlankLinks: this.getFeatureSettingEnabled('trimBlankLinks', 'enabled'),
+            }).trim();
+            this.log.info('Content markdown', result, root);
+            return result;
+        };
+
+        if (contentRoot) {
+            content += extractContent(contentRoot);
+        }
+        // If the main content is empty, use the body
+        if (content.length === 0 && contentRoot !== document.body && this.getFeatureSettingEnabled('bodyFallback', 'enabled')) {
+            contentRoot = document.body;
+            content += extractContent(contentRoot);
         }
-        content = content.trim();
 
         // Store the full content length before truncation
         this.fullContentLength = content.length;

injected/unit-test/fixtures/page-context/README.md

@@ -0,0 +1,53 @@
+# Page Context DOM-to-Markdown Tests
+
+This directory contains test fixtures for testing the `domToMarkdown` function from `page-context.js`.
+
+## Directory Structure
+
+- `output/` - Generated markdown files from test runs (temporary, regenerated on each run)
+- `expected/` - Expected markdown output files (committed to git)
+
+## How It Works
+
+The test suite (`page-context-dom.spec.js`) does the following:
+
+1. **Creates test cases** with HTML snippets and settings for `domToMarkdown`
+2. **Converts HTML to Markdown** using JSDom to simulate a browser environment
+3. **Writes output** to `output/` directory for inspection
+4. **Compares output** with expected files in `expected/` directory
+5. **Fails if different** - Any difference between output and expected causes test failure
+
+## Test Cases
+
+The suite includes 20 test cases covering:
+
+- Basic HTML elements (paragraphs, headings, lists, links, images)
+- Formatting (bold, italic, mixed formatting)
+- Complex structures (nested lists, articles, blog posts)
+- Edge cases (hidden content, empty links, whitespace handling)
+- Configuration options (max length truncation, excluded selectors, trim blank links)
+
+## Updating Expected Output
+
+When the `domToMarkdown` function behavior changes:
+
+1. Review the changes in `output/` directory
+2. If changes are correct, copy them to `expected/`:
+   ```bash
+   cp unit-test/fixtures/page-context/output/*.md unit-test/fixtures/page-context/expected/
+   ```
+3. Commit the updated expected files
+
+## Running Tests
+
+```bash
+npm run test-unit -- unit-test/page-context-dom.spec.js
+```
+
+## Why This Approach?
+
+- **Visibility**: Output files make it easy to review markdown generation
+- **Regression detection**: Tests fail on any unintended changes
+- **Documentation**: Expected files serve as examples of the function's behavior
+- **Easy updates**: Simple to update baselines when behavior intentionally changes
+

injected/unit-test/fixtures/page-context/expected/article-structure.md

@@ -0,0 +1,15 @@
+# Article Title
+ By **Author Name**
+ This is the introduction paragraph with some *emphasis*.
+
+## First Section
+ Content of the first section.
+
+
+- Point one
+
+- Point two
+
+
+## Second Section
+ Content with a [link](https://example.com).

injected/unit-test/fixtures/page-context/expected/blog-post.md

@@ -0,0 +1,16 @@
+# Blog Post Title
+ Published on January 1, 2024
+
+![Header image](header.jpg)
+ Lorem ipsum dolor sit amet, consectetur adipiscing elit.
+
+## Key Takeaways
+
+
+- First takeaway
+
+- Second takeaway
+
+- Third takeaway
+
+ Read more on [our blog](https://blog.example.com).

injected/unit-test/fixtures/page-context/expected/bold-and-italic.md

@@ -0,0 +1 @@
+This is **bold** and this is *italic*.

injected/unit-test/fixtures/page-context/expected/complex-nested.md

@@ -0,0 +1,5 @@
+# Article Title
+Introduction paragraph.
+
+## Section 1
+Section content with **bold** text.

injected/unit-test/fixtures/page-context/expected/empty-link-without-trim.md

@@ -0,0 +1 @@
+[](https://example.com)

injected/unit-test/fixtures/page-context/expected/excluded-selectors.md

@@ -0,0 +1,2 @@
+Keep this
+Keep this too

injected/unit-test/fixtures/page-context/expected/headings.md

@@ -0,0 +1,5 @@
+# Main Heading
+
+## Subheading
+
+### Sub-subheading

injected/unit-test/fixtures/page-context/expected/hidden-content.md

@@ -0,0 +1 @@
+Visible text

injected/unit-test/fixtures/page-context/expected/image.md

@@ -0,0 +1 @@
+![A beautiful landscape](photo.jpg)

injected/unit-test/fixtures/page-context/expected/line-breaks.md

@@ -0,0 +1,3 @@
+First line
+Second line
+Third line

injected/unit-test/fixtures/page-context/expected/links.md

@@ -0,0 +1 @@
+Visit [our website](https://example.com) for more info.

injected/unit-test/fixtures/page-context/expected/max-length-truncation.md

@@ -0,0 +1 @@
+This is a very long paragraph ...

injected/unit-test/fixtures/page-context/expected/mixed-formatting.md

@@ -0,0 +1 @@
+This has ***bold and italic*** together.

injected/unit-test/fixtures/page-context/expected/multiple-paragraphs.md

@@ -0,0 +1,3 @@
+First paragraph.
+Second paragraph.
+Third paragraph.

injected/unit-test/fixtures/page-context/expected/nested-lists.md

@@ -0,0 +1,3 @@
+- Item 1 - Subitem 1.1 - Subitem 1.2
+
+- Item 2

injected/unit-test/fixtures/page-context/expected/ordered-list.md

@@ -0,0 +1,5 @@
+- First step
+
+- Second step
+
+- Third step

injected/unit-test/fixtures/page-context/expected/simple-paragraph.md

@@ -0,0 +1 @@
+This is a simple paragraph.

injected/unit-test/fixtures/page-context/expected/unordered-list.md

@@ -0,0 +1,5 @@
+- First item
+
+- Second item
+
+- Third item

injected/unit-test/fixtures/page-context/expected/whitespace-handling.md

@@ -0,0 +1 @@
+Text with multiple spaces