feat(metadata): enhance metadata handling by sanitizing extracted data and updating documentation for file storage

internal/metadata/normalize.go

@@ -110,6 +110,15 @@ func MarkMetadataComplete(base thoughttypes.ThoughtMetadata, capture config.Capt
 	return out
 }
 
+func SanitizeExtracted(in thoughttypes.ThoughtMetadata) thoughttypes.ThoughtMetadata {
+	in.Attachments = nil
+	in.MetadataStatus = ""
+	in.MetadataUpdatedAt = ""
+	in.MetadataLastAttemptedAt = ""
+	in.MetadataError = ""
+	return in
+}
+
 func normalizeList(values []string, limit int) []string {
 	seen := make(map[string]struct{}, len(values))
 	result := make([]string, 0, len(values))

internal/metadata/normalize_test.go

@@ -107,6 +107,38 @@ func TestNormalizeDedupesAttachmentsByFileID(t *testing.T) {
 	}
 }
 
+func TestSanitizeExtractedDropsAttachmentsAndMetadataControlFields(t *testing.T) {
+	id := uuid.New()
+
+	got := SanitizeExtracted(thoughttypes.ThoughtMetadata{
+		Type:                    "idea",
+		Attachments:             []thoughttypes.ThoughtAttachment{{FileID: id, Name: "secret.pdf"}},
+		MetadataStatus:          MetadataStatusFailed,
+		MetadataUpdatedAt:       "2026-03-30T10:00:00Z",
+		MetadataLastAttemptedAt: "2026-03-30T10:01:00Z",
+		MetadataError:           "boom",
+	})
+
+	if len(got.Attachments) != 0 {
+		t.Fatalf("Attachments len = %d, want 0", len(got.Attachments))
+	}
+	if got.MetadataStatus != "" {
+		t.Fatalf("MetadataStatus = %q, want empty", got.MetadataStatus)
+	}
+	if got.MetadataUpdatedAt != "" {
+		t.Fatalf("MetadataUpdatedAt = %q, want empty", got.MetadataUpdatedAt)
+	}
+	if got.MetadataLastAttemptedAt != "" {
+		t.Fatalf("MetadataLastAttemptedAt = %q, want empty", got.MetadataLastAttemptedAt)
+	}
+	if got.MetadataError != "" {
+		t.Fatalf("MetadataError = %q, want empty", got.MetadataError)
+	}
+	if got.Type != "idea" {
+		t.Fatalf("Type = %q, want idea", got.Type)
+	}
+}
+
 func TestMarkMetadataPendingTracksAttemptWithoutClearingPreviousSuccess(t *testing.T) {
 	attempt := time.Date(2026, 3, 30, 10, 0, 0, 0, time.UTC)
 	base := thoughttypes.ThoughtMetadata{