feat(capture): extract page main content as Markdown

Clipping a page with nothing selected captured only its title, URL, and
a
one-line description, which is closer to a bookmark than a clip.

- Prefill the editor with the page's article body as Markdown,
  preserving
  headings, lists, code blocks, and tables
- Run extraction in the popup against an inert document, so the injected
  capture stays self-contained and the extractor is testable without a
  tab
- Load the extractor lazily, leaving the popup's initial parse unchanged
- Keep an explicit selection ahead of the extracted article
- Drop images from extracted articles rather than hotlinking them, since
  images are uploaded as attachments; figure captions are kept
- Explain the fallback when a page has no article to extract

Author: boojack

README.md

@@ -27,7 +27,7 @@ To verify a download on Linux, run `sha256sum --ignore-missing -c SHA256SUMS` in
 
 ## Features
 
-- Capture the page title, URL, and readable description.
+- Capture a page's main content as Markdown, with its title, URL, and readable description.
 - Convert selected text, links, lists, code, and tables to Markdown.
 - Upload selected images as Memos attachments.
 - Review and edit every clip before saving.
@@ -39,6 +39,14 @@ To verify a download on Linux, run `sha256sum --ignore-missing -c SHA256SUMS` in
 
 The default template keeps the captured content first and adds a link back to the source. Empty fields are removed automatically.
 
+## What gets captured
+
+Selecting text before you clip captures exactly that selection, quoted, along with any images in it.
+
+With nothing selected, the clipper extracts the page's main content — the article body, without navigation, sidebars, or footers — and converts it to Markdown, preserving headings, lists, code blocks, and tables. Images in an extracted article are left out rather than linked, so a clip never depends on someone else's server staying up; select an image, or the region containing it, to save it as an attachment.
+
+Pages with no article to extract, such as dashboards, search results, and feeds, fall back to the page's own summary and link. The popup says which of these happened, and every clip is editable before you save it.
+
 ## How to use
 
 1. Open the extension settings and choose a connection method:

package.json

@@ -27,6 +27,7 @@
     "@fontsource-variable/geist-mono": "^5.2.8",
     "class-variance-authority": "^0.7.1",
     "clsx": "^2.1.1",
+    "defuddle": "^0.19.2",
     "lucide-react": "^1.22.0",
     "next-themes": "^0.4.6",
     "react": "^19.2.7",

pnpm-lock.yaml

@@ -152,6 +152,7 @@
   "popupCaptureNoDescriptionSelection": {
     "message": "Auswahl und Seitenlink wurden erfasst – keine lesbare Beschreibung gefunden."
   },
+  "popupCaptureNoArticle": { "message": "Seitenbeschreibung und Seitenlink wurden erfasst – kein lesbarer Artikel gefunden." },
   "popupCaptureNoDescription": {
     "message": "Nur der Seitenlink wurde erfasst – keine lesbare Beschreibung gefunden."
   },