Caddy’s stripHTML template function cannot reliably remove all HTML tags from input strings. Certain malformed HTML, such as <<>img src=x onerror=alert()>, can bypass the tag-stripping logic, potentially leaving dangerous content in the output if it is later rendered as HTML. This may allow client-side XSS in cases where untrusted strings are rendered unsafely.
The vulnerability originates from funcStripHTML in:
caddy/caddy/caddyhttp/templates/tplcontext.go
func (TemplateContext) funcStripHTML(s string) string {
var buf bytes.Buffer
var inTag, inQuotes bool
var tagStart int
for i, ch := range s {
if inTag {
if ch == '>' && !inQuotes {
inTag = false
} else if ch == '<' && !inQuotes {
// false start
buf.WriteString(s[tagStart:i])
tagStart = i
} else if ch == '"' {
inQuotes = !inQuotes
}
continue
}
if ch == '<' {
inTag = true
tagStart = i
continue
}
buf.WriteRune(ch)
}
if inTag {
// false start
buf.WriteString(s[tagStart:])
}
return buf.String()
}
Caddyfile setup
:8080 {
root * ./site
file_server
templates
}
Template file (index.html)
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="UTF-8">
<title>StripHTML Bypass Test</title>
</head>
<body>
<p>{{ stripHTML "<<>img src=x onerror=alert('XSS')>" }}</p>
</body>
</html>
The payload exploits the false start branch to smuggle a literal < back into the output, then uses the following > to terminate the parser’s tag state, leaving a valid <img ...> tag behind.
Tested in v2.11.3
Malformed HTML can bypass stripHTML, potentially allowing arbitrary HTML or JavaScript to be rendered if the output is used unsafely, leading to client-side XSS.
AI assisted in writing the report description; however, the discovery of the issue has been done manually.
{
"github_reviewed": true,
"github_reviewed_at": "2026-06-16T21:28:55Z",
"nvd_published_at": null,
"severity": "MODERATE",
"cwe_ids": [
"CWE-116"
]
}