
π₯ XSS2Shell β CVE-2026-64638 Scanner & PoC Toolkit
Behavior-first mass scanner and evidence-grade PoC generator for the WordPress
pre-auth XSS-to-RCE chain affecting 500M+ websites.
Detection only. No weaponization. Built for bug bounty programs and blue teams.
What is this? β’ Quick Start β’ Shodan Dorks β’ Usage β’ Decision Matrix β’ Detection β’ FAQ
Hunt for potentially vulnerable WordPress instances across the internet before scanning:
http.component:"wordpress" -http.title:"Just a moment"
Finds WordPress sites while excluding Cloudflare "I'm Under Attack" mode / bot-protection pages that will block or challenge automated requests.
http.component:"wordpress" http.title:"Log In"
Returns only WordPress login pages β the exact attack surface for CVE-2026-64638.
http.component:"wordpress" "wp-content" "?ver=7.0" -"?ver=7.0.3"
Flags WordPress 7.0.x instances without the 7.0.3 patch by asset version fingerprinting.
http.component:"wordpress" http.html:"wp-login.php"
Catches sites where wp-login.php is reachable but may not be the current page β broader coverage.
http.component:"wordpress" -http.title:"Just a moment" -http.title:"Attention Required" -org:"Cloudflare"
Aggressive filter that strips out most Cloudflare-fronted targets. Use when scanning at scale with --active β Cloudflare will rate-limit or block the probe request.
Tip: Export Shodan results with
shodan downloadand pipe the hostnames directly intoxss2shell_mass.py -i.
On August 7, 2026, pwn.ai disclosed CVE-2026-64638 (XSS2Shell) β a critical pre-authentication cross-site scripting vulnerability in WordPress Core that chains all the way to remote code execution on the server. [citation:pwn.ai blog]
The bug exploits a parser disagreement between PHP's strip_tags() and WordPress's wp_kses_post():
strip_tags() uses < immediately followed by a letter to identify HTML tags. < area id=...> (with a space) is treated as text β it survives.wp_kses_post() (KSES) recognizes < area as a valid <area> element β and <area> is allowlisted in KSES. [citation:pwn.ai blog]One failed login with a specially crafted username < area id=ajaxurl href=/?rest_route=/&_method=GET&_jsonp=alert>... bypasses both sanitizers, gets rendered as live DOM on the login page, hijacks WordPress's own user-profile.js script via DOM clobbering, and fires alert() in the WordPress origin β zero clicks, zero authentication, zero cookies required. [citation:pwn.ai blog]
Escalated to a logged-in administrator? The same primitive steals Application Passwords via Same Origin Method Execution (SOME), uploads a malicious plugin, and executes PHP as www-data. [citation:pwn.ai blog] [citation:hadrian.io blog]
Affected: WordPress 6.4 through 7.0.2 β patched in 7.0.3 with backports to 4.7+.
Impact: ~500 million websites at time of disclosure. [citation:pwn.ai blog]
| Resource | Link |
|---|---|
| Original Disclosure (pwn.ai) | pwn.ai/blog/xss2shell |
| Hadrian Technical Analysis | hadrian.io/blog/wordpress-xss2shell |
| WordPress Advisory (GHSA) | GHSA-52p2-r8wf-jcrf |
| SOME Attack Research (2022) | pwn.ai/blog/bypass-csp-using-wordpress |
| WordPress 7.0.3 Release | wordpress.org/news/2026/08/wordpress-7-0-3-release |
This is a detection-only toolkit. It does not weaponize the vulnerability β it gives security researchers, bug bounty hunters, and blue teams everything needed to:
"A version string says what patch level the code should be.
Only the login-page sanitizer behavior says whether the bug fires."
Managed hosts silently backport security patches without bumping version strings. Login-hardening plugins replace the error message entirely, killing the reflection channel even on insecure versions. Version-only scanners produce false positives and false negatives. This scanner sends a single benign probe and classifies the actual sanitizer behavior.
git clone https://github.com/jakestone/xss2shell.git
cd xss2shell
pip install -r requirements.txt
# Passive β no probes sent to target, version + endpoint fingerprinting only
python3 xss2shell_mass.py -i domains.txt -o results
# Active β sends ONE benign failed-login per host (authorized assets only!)
python3 xss2shell_mass.py -i domains.txt -o results --active --workers 80
# Single target
python3 make_poc.py --target https://blog.example.com
# Batch from scanner output
python3 make_poc.py --from-results results.csv -o pocs/
Open the generated .poc.html in your browser while recording video β if alert() fires, you've captured pre-auth XSS evidence.
xss2shell_mass.py)usage: xss2shell_mass.py [-h] -i INPUT [-o OUTPUT]
[--active] [--workers WORKERS]
[--timeout TIMEOUT] [--quiet]
| Flag | Description |
|---|---|
-i, --input | File with one host per line (bare domain or full URL) |
-o, --output | Base path for output files (generates .csv + .json) |
--active | Enable behavioral probe β one failed login per host |
--workers | Thread pool size (default: 50, max ~200 for good connections) |
--timeout | HTTP timeout in seconds (default: 10) |
--quiet | Only print confirmed_vulnerable, vulnerable, and likely_vulnerable |
?ver= params, wp-content references)user-profile.js gadget enqueued, core asset versions/?rest_route=/&_method=GET&_jsonp=<random> β is the JSONP pathway open?--active flag)