This vector attempts to detect tags that consume markup. Then shows select can be used to break out of the consumption. This is interesting because tags you would normally expect could be closed by itself can actually be closed by a closing select e.g. <select><noscript></select><img src onerror=alert(1)>. Note plaintext also works but this breaks Shazzer. <select><plaintext></select><img src onerror=alert(1)> Note: Apple have since fixed this issue with Safari
<script>
window.scriptsExecuted = {};
function countExecution(tag) {
if(!scriptsExecuted[tag]) {
scriptsExecuted[tag] = 0;
}
scriptsExecuted[tag]++;
}
</script><$[data1]><script>countExecution('$[data1]')</script></$[data1]>
<select><$[data1]></select><script>countExecution('$[data1]')</script></$[data1]></select>for (const [tag, amount] of Object.entries(scriptsExecuted)) {
if(amount === 1) {
log(tag)
}
}<iframe><script>countExecution('iframe')</script></iframe>
<select><iframe></select><script>countExecution('iframe')</script></iframe></select><noembed><script>countExecution('noembed')</script></noembed>
<select><noembed></select><script>countExecution('noembed')</script></noembed></select><noframes><script>countExecution('noframes')</script></noframes>
<select><noframes></select><script>countExecution('noframes')</script></noframes></select><noscript><script>countExecution('noscript')</script></noscript>
<select><noscript></select><script>countExecution('noscript')</script></noscript></select><style><script>countExecution('style')</script></style>
<select><style></select><script>countExecution('style')</script></style></select><title><script>countExecution('title')</script></title>
<select><title></select><script>countExecution('title')</script></title></select><xmp><script>countExecution('xmp')</script></xmp>
<select><xmp></select><script>countExecution('xmp')</script></xmp></select>Chrome 158.0.8127.37 ran this vector and matched nothing.
Firefox 159.0 ran this vector and matched nothing.
Microsoft Edge 157.0.4341.46 ran this vector and matched nothing.
Safari 27.2 ran this vector and matched nothing.
Safari 27.0 could not complete this vector: Safari 27.0 reported no matches, but Safari 26.6.2 has results; held until a second peer reports the same