also guard against invalid UTF-8 in input

app/services/import/jira_wiki_markup/parser.rb

@@ -46,10 +46,16 @@ module Import
       )
 
       def initialize(text)
-        @text = text.dup
+        # Normalize any input into a safe, mutable UTF-8 string: nil becomes "",
+        # and invalid byte sequences are dropped so downstream regex/StringScanner
+        # operations cannot raise ArgumentError on malformed input.
+        @text = text.to_s.dup
+        @text.scrub!("") unless @text.valid_encoding?
       end
 
       def parse
+        return N::Document.new(children: []) if @text.blank?
+
         preprocess
         blocks = parse_blocks
         N::Document.new(children: blocks)

app/services/import/jira_wiki_markup_converter.rb

@@ -35,8 +35,6 @@ module Import
     end
 
     def convert
-      return "" if @text.blank?
-
       ast = JiraWikiMarkup::Parser.new(@text).parse
       JiraWikiMarkup::Renderer.new(ast).render
     end

app/workers/import/jira_fetch_and_import_projects_job.rb

@@ -102,8 +102,6 @@ module Import
     end
 
     def collect_markup_mentions(text, mention_usernames)
-      return if text.blank?
-
       ast = JiraWikiMarkup::Parser.new(text).parse
       collect_mentions_from_node(ast, mention_usernames)
     end

spec/services/import/jira_wiki_markup_converter_spec.rb

@@ -51,6 +51,38 @@ RSpec.describe Import::JiraWikiMarkupConverter do
 
       it { is_expected.to eq("This is not {code} and not [a link]") }
     end
+
+    context "with invalid UTF-8 byte sequences in the input" do
+      it "drops a stray invalid byte and keeps the surrounding text" do
+        input = "Hello \xFF world".dup
+        expect(input.valid_encoding?).to be(false)
+        expect(described_class.new(input).convert).to eq("Hello  world")
+      end
+
+      it "drops a stray continuation byte" do
+        input = "abc \x80 def".dup
+        expect(input.valid_encoding?).to be(false)
+        expect(described_class.new(input).convert).to eq("abc  def")
+      end
+
+      it "drops a truncated multi-byte sequence" do
+        input = "pre \xC3 post".dup
+        expect(input.valid_encoding?).to be(false)
+        expect(described_class.new(input).convert).to eq("pre  post")
+      end
+
+      it "preserves valid multi-byte characters while dropping only the invalid byte" do
+        input = "héllo \xFF world".dup
+        expect(input.valid_encoding?).to be(false)
+        expect(described_class.new(input).convert).to eq("héllo  world")
+      end
+
+      it "still parses formatting around invalid bytes inside delimiters" do
+        input = "*bold\xFFtext*".dup
+        expect(input.valid_encoding?).to be(false)
+        expect(described_class.new(input).convert).to eq("**boldtext**")
+      end
+    end
   end
 
   describe "line ending normalization" do