Check a live site from the outside
phpwarden probe looks at a running PHP or WordPress site the way a visitor would and tells you what is exposed, ranked by what attackers are using. It reads; it never attacks.
What it tells you
- What is running, and whether it is being attacked. The WordPress, plugin, theme and PHP versions the site reveals are matched against the phpwarden feed and ranked the same way as every other report: known exploited (CISA KEV) first, then EPSS, then severity.
- Files that should not be public.
.env, a readable.gitdirectory,wp-config.phpbackups, database dumps,debug.log,phpinfo.php. - Debug and admin endpoints left open. Laravel Ignition, Telescope and Horizon, the Symfony profiler, Adminer, phpMyAdmin, XML-RPC, the WordPress user list.
- Transport and header basics. HTTPS redirect, HSTS, CSP, cookie flags, certificate expiry, directory listing in uploads.
$ phpwarden probe https://client.example
phpwarden probe 2.4.0 -- https://client.example
mode: active ownership: verified (well-known) requests: 33
fingerprint: WordPress 6.4.2; 1 plugins; frameworks: wordpress
HIGH [EXPLOITED] WordPress core 6.4.2: unauthenticated local .php file inclusion ... CVE-2026-87902
CRITICAL [EXPLOITED] plugin litespeed-cache 6.2: Unauthenticated privilege escalation ... CVE-2024-28000
CRITICAL Exposed .env file -- 3 variables exposed (names only): APP_KEY, DB_PASSWORD, APP_DEBUG
CRITICAL Exposed .git repository -- repository HEAD readable (ref: ...)
MEDIUM WordPress XML-RPC enabled -- XML-RPC enabled
LOW Missing Content-Security-Policy -- no CSP header
...
summary: 3 critical, 4 high, 4 medium, 4 low, 1 info; 3 known exploited
That is real output from a test site, shortened. Secrets are never copied into a report: an exposed
.env is listed by variable name only.
For agencies and hosts
You look after sites you did not build and cannot always log in to. The probe needs nothing installed on the site: a URL and proof that you are allowed to check it. Run it when you take on a client, and on a schedule after that. It answers the question a client actually asks, "is anything on my site being attacked right now?", from the outside, in under a minute.
It works alongside the checks that run on the server. phpwarden wordpress and the WordPress plugin
see every installed component; the probe sees only what the site shows a visitor. Use the probe where you have
no server access, and both where you do.
The full check only runs on sites you can prove are yours
A tool that anyone can point at any site is a tool for attackers. So the active checks run only after you publish a token on the site, in one of three ways:
- a DNS TXT record:
phpwarden-verify=<token> - a file at
/.well-known/phpwarden-verify.txt - a
<meta name="phpwarden-verify">tag on the homepage
phpwarden probe https://client.example --print-token prints the token and the exact text to
publish. On a site that has not been verified, the probe requests the homepage, looks for the ownership file,
and stops.
What it never does
- It sends GET and HEAD requests only. No POST, no payloads, no login attempts, no brute force. (The one exception is the optional signed-in check you turn on yourself: one login POST to your own verified site.)
- It requests a fixed list of well-known paths. It does not crawl.
- It makes at most 60 requests per run by default, with a pause between them, and stops after 90 seconds.
- It does not follow a redirect to another hostname.
- It identifies itself:
phpwarden-probe/<version> (+https://phpwarden.io/probe).
If you found this page in your server logs
A request with the User-Agent phpwarden-probe came from someone running the open-source phpwarden
CLI against your site. If your site has not been verified by them, it received at most two requests: the
homepage and /.well-known/phpwarden-verify.txt. To report misuse, write to
[email protected].
Limits
- It sees only what the site reveals. A site that hides its versions gives the probe less to match.
- A PHP version read from a header cannot tell which SAPI or operating system is in use, so engine findings that depend on those are marked "(possibly)".
- It is not a web application scanner. It does not test your forms for SQL injection or XSS; use a tool built for that, such as OWASP ZAP.