fix: Properly handle CDATA sections in XML output

This commit fixes an issue where the XML export would fail with
"unexpected EOF in CDATA section" errors when file content contained
the CDATA end marker sequence ']]>'.

The fix implements a proper CDATA handling strategy that:
- Detects all occurrences of ']]>' in file content
- Splits the content around these markers
- Creates properly nested CDATA sections to preserve the original content
- Ensures all XML output is well-formed regardless of source content

This approach maintains the efficiency of CDATA for storing large code
blocks while ensuring compatibility with all possible file content.

Fixes the XML validation error that would occur when processing files
containing CDATA end marker sequences.

Author: imertz

chat.json

@@ -214,11 +214,12 @@ func OutputGitRepo(repo *GitRepo, preambleFile string, scrubComments bool) (stri
 	return output, nil
 }
 
+
 func OutputGitRepoXML(repo *GitRepo, scrubComments bool) (string, error) {
 	if scrubComments {
-			for i, file := range repo.Files {
-					repo.Files[i].Contents = utils.RemoveCodeComments(file.Contents)
-			}
+		for i, file := range repo.Files {
+			repo.Files[i].Contents = utils.RemoveCodeComments(file.Contents)
+		}
 	}
 	var result strings.Builder
 	result.WriteString("<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n")
@@ -229,16 +230,38 @@ func OutputGitRepoXML(repo *GitRepo, scrubComments bool) (string, error) {
 	result.WriteString("    <files>\n")
 
 	for _, file := range repo.Files {
-			result.WriteString("        <file>\n")
-			result.WriteString(fmt.Sprintf("            <path>%s</path>\n", escapeXML(file.Path)))
-			result.WriteString(fmt.Sprintf("            <tokens>%d</tokens>\n", file.Tokens))
+		result.WriteString("        <file>\n")
+		result.WriteString(fmt.Sprintf("            <path>%s</path>\n", escapeXML(file.Path)))
+		result.WriteString(fmt.Sprintf("            <tokens>%d</tokens>\n", file.Tokens))
+		
+		// Split content around CDATA end marker (]]>) and create multiple CDATA sections
+		contents := file.Contents
+		result.WriteString("            <contents>")
+		
+		for {
+			idx := strings.Index(contents, "]]>")
+			if idx == -1 {
+				// No more CDATA end markers, write remaining content in one CDATA section
+				result.WriteString("<![CDATA[")
+				result.WriteString(contents)
+				result.WriteString("]]>")
+				break
+			}
+			
+			// Write content up to the CDATA end marker
+			result.WriteString("<![CDATA[")
+			result.WriteString(contents[:idx+2]) // Include the "]]" part
+			result.WriteString("]]>") // Close this CDATA section
 
-			safeContents := strings.ReplaceAll(file.Contents, "]]]]><![CDATA[>", "]]]]]]><![CDATA[><![CDATA[>")
+			// Start a new CDATA section with the ">" character
+			result.WriteString("<![CDATA[>")
 
-			result.WriteString("            <contents><![CDATA[")
-			result.WriteString(safeContents)
-			result.WriteString("]]]]><![CDATA[></contents>\n")
-			result.WriteString("        </file>\n")
+			// Move past the "]]>" in the original content
+			contents = contents[idx+3:]
+		}
+		
+		result.WriteString("</contents>\n")
+		result.WriteString("        </file>\n")
 	}
 
 	result.WriteString("    </files>\n")
@@ -250,10 +273,10 @@ func OutputGitRepoXML(repo *GitRepo, scrubComments bool) (string, error) {
 	repo.TotalTokens = tokenCount
 
 	outputStr = strings.Replace(
-			outputStr, 
-			"<total_tokens>PLACEHOLDER</total_tokens>", 
-			fmt.Sprintf("<total_tokens>%d</total_tokens>", tokenCount), 
-			1,
+		outputStr, 
+		"<total_tokens>PLACEHOLDER</total_tokens>", 
+		fmt.Sprintf("<total_tokens>%d</total_tokens>", tokenCount), 
+		1,
 	)
 
 	return outputStr, nil