Add source attachment metadata compatibility

docs/pdf-adapter.md

@@ -23,7 +23,7 @@ native system services, or renderer-specific tooling.
 - Scanned or image-only PDFs that require OCR.
 - Encrypted or permission-restricted PDFs.
 - Pixel-perfect layout reconstruction.
-- Table, figure, annotation, form, signature, and attachment extraction.
+- Table, figure, annotation, form, signature, and rich attachment extraction.
 - PDF writing/export.
 
 ## Options
@@ -43,3 +43,9 @@ and originating PDF page object id.
 Quality metadata records the extraction backend, document page count, selected
 pages, extracted page count, page coverage, skipped pages, warning count,
 lossiness, and confidence.
+
+`NormalizedMarkdownDocument.attachments` may include read-side metadata for
+embedded file streams and image-resource signals when the stdlib parser can
+detect them. Embedded files include byte size and digest. Image resources are
+signal-only descriptors with page/object provenance; the adapter does not
+extract image bytes or perform OCR.