Input sanitization & validation

SkillCommunication

Use when reading any external input in WordPress — $_GET, $_POST, $_REQUEST, $_COOKIE, REST params, shortcode/block attributes, option/meta values, or remote API responses. Unslashes then sanitizes to the correct type (sanitize_text_field, sanitize_email, absint, sanitize_key, wp_kses_post, esc_url_raw) and validates values against expected sets. Apply proactively before storing or using any untrusted value.

Available today. Use it from your connected AI after setup.

Connect ahel once, and every AI you use reads what you have installed.

Then ask your AI: use the Input sanitization & validation skill

What this skill tells your AI

The instructions your AI receives, as published by wpultimatesecurity/wordpress-security-skills in skills/input-sanitization-validation/SKILL.md and read by ahel’s review.

When to use this skill

Use this skill whenever a value enters your code from outside:

  • Superglobals: $_GET, $_POST, $_REQUEST, $_COOKIE, $_SERVER.
  • REST API parameters, AJAX payloads, form fields.
  • Shortcode attributes, block attributes, widget settings.
  • Values read back from get_option, post meta, or user meta that originated from users.
  • Bodies returned by remote APIs (wp_remote_get, etc.).

Sanitize on input, escape on output — two different jobs. This skill covers input; see output-escaping for output.

Core principles (and why they matter)

  1. All external input is untrusted. Even data from your own admin screens can be forged (CSRF) or tampered with. Sanitize regardless of source.
  2. Unslash before sanitizing. WordPress adds slashes to superglobals. Calling wp_unslash() first prevents corrupted data (it\'s instead of it's) and ensures sanitizers see the real value.
  3. Sanitize to the expected type. A function that expects an integer should call absint(); an email, sanitize_email(). Type-appropriate sanitizing shrinks the attack surface far more than a generic catch-all.
  4. Validation ≠ sanitization. Sanitizing cleans a value; validating rejects values that don't belong. Use both: sanitize the shape, then validate against allowed sets (in_array with strict comparison, range checks, is_email).
  5. Prefer allowlists over blocklists. Define what is permitted, reject the rest.
  6. Sanitize rich content with wp_kses/wp_kses_post, never with strip_tagsstrip_tags leaves dangerous attributes and is not XSS-safe.

Step-by-step implementation

  1. Confirm the value is set with isset() / null coalescing and a safe default.
  2. Unslash: wp_unslash( $_POST['x'] ).
  3. Sanitize to type: pick the function matching the data (see cheatsheet).
  4. Validate: check ranges, allowed values, formats; reject invalid input.
  5. Use or store the clean value. Re-escape when later output.
$status   = isset( $_POST['status'] ) ? sanitize_key( wp_unslash( $_POST['status'] ) ) : '';
$allowed  = array( 'draft', 'publish', 'pending' );
$status   = in_array( $status, $allowed, true ) ? $status : 'draft'; // validate + fallback

Supporting references

ReferenceLoad when
Input sanitization & validation checklistBefore final verification of the input sanitization & validation controls.
Sanitization & validation cheatsheetChoosing the applicable WordPress API or control for input sanitization and validation.

Common AI mistakes / anti-patterns

Mistake 1 — Sanitizing without unslashing

// ❌ Insecure/buggy: slashes corrupt the value and bypass intent.
$name = sanitize_text_field( $_POST['name'] );
// ✅ Correct: unslash first.
$name = isset( $_POST['name'] ) ? sanitize_text_field( wp_unslash( $_POST['name'] ) ) : '';

Mistake 2 — Using strip_tags() as if it were XSS-safe

// ❌ Insecure: strip_tags keeps event handlers/attributes and is not escaping.
$bio = strip_tags( $_POST['bio'] );
update_user_meta( $user_id, 'bio', $bio );
// ✅ Secure: allow a known-safe subset with wp_kses_post (or a custom allowlist).
$bio = isset( $_POST['bio'] ) ? wp_kses_post( wp_unslash( $_POST['bio'] ) ) : '';
update_user_meta( $user_id, 'bio', $bio );

Mistake 3 — Wrong sanitizer for the type

// ❌ Insecure: text sanitizer on something that must be an integer / URL.
$post_id = sanitize_text_field( $_GET['post_id'] ); // "12 OR 1=1" survives as a string
$link    = sanitize_text_field( $_POST['link'] );   // allows javascript: scheme
// ✅ Secure: type-correct sanitizers.
$post_id = absint( $_GET['post_id'] ?? 0 );
$link    = esc_url_raw( wp_unslash( $_POST['link'] ?? '' ) ); // for storage; esc_url for output

Mistake 4 — Sanitizing but never validating

// ❌ Weak: clean string, but any value is accepted.
$role = sanitize_key( wp_unslash( $_POST['role'] ?? '' ) );
$user->set_role( $role ); // attacker sends "administrator"
// ✅ Secure: validate against an allowlist.
$role    = sanitize_key( wp_unslash( $_POST['role'] ?? '' ) );
$allowed = array( 'subscriber', 'contributor', 'author' );
if ( ! in_array( $role, $allowed, true ) ) {
    wp_die( esc_html__( 'Invalid role.', 'my-plugin' ), 400 );
}
$user->set_role( $role );

Mistake 5 — Sanitizing arrays as if they were scalars

// ❌ Buggy/insecure: passing an array to a scalar sanitizer.
$ids = sanitize_text_field( $_POST['ids'] ); // returns "" for arrays
// ✅ Correct: map the sanitizer across the array.
$ids = isset( $_POST['ids'] ) && is_array( $_POST['ids'] )
    ? array_map( 'absint', wp_unslash( $_POST['ids'] ) )
    : array();

Mistake 6 — Open redirect via unvalidated URL

// ❌ Insecure: attacker can redirect to a phishing site.
$redirect = $_GET['redirect_to'];
wp_redirect( $redirect );
exit;
// ✅ Secure: validate redirect targets with wp_validate_redirect and use wp_safe_redirect.
$redirect = isset( $_GET['redirect_to'] ) ? esc_url_raw( wp_unslash( $_GET['redirect_to'] ) ) : '';
$redirect = wp_validate_redirect( $redirect, admin_url() );
wp_safe_redirect( $redirect );
exit;

Mistake 7 — Trusting the whole request array (mass assignment)

// ❌ Insecure: every client-supplied field is written through — including ones
// the form never rendered. A forged POST adds post_author or post_status=publish.
wp_update_post( wp_unslash( $_POST['post'] ) );

foreach ( $_POST['settings'] as $key => $value ) {
    update_user_meta( $user_id, 'my_plugin_' . sanitize_key( $key ), $value );
}
// Inside an authenticated admin-post handler; the form uses
// wp_nonce_field( 'my_plugin_edit_post_' . $post_id ).
$raw = $_POST['post'] ?? null;
if ( ! is_array( $raw )
    || ! isset( $raw['ID'], $raw['post_title'], $raw['post_content'] )
    || ! is_string( $raw['ID'] )
    || ! is_string( $raw['post_title'] )
    || ! is_string( $raw['post_content'] )
) {
    wp_die( 'Invalid payload.', '', array( 'response' => 400 ) );
}
$post_id = filter_var( wp_unslash( $raw['ID'] ), FILTER_VALIDATE_INT, array(
    'options' => array( 'min_range' => 1 ),
) );
if ( false === $post_id ) {
    wp_die( 'Invalid post ID.', '', array( 'response' => 400 ) );
}
check_admin_referer( 'my_plugin_edit_post_' . $post_id );
$post = get_post( $post_id );
if ( ! $post || 'post' !== $post->post_type || ! current_user_can( 'edit_post', $post_id ) ) {
    wp_die( 'Forbidden.', '', array( 'response' => 403 ) );
}
// Build only the allowed fields. Never merge the request back into this array.
$data = array(
    'ID'           => $post_id,
    'post_title'   => sanitize_text_field( wp_unslash( $raw['post_title'] ) ),
    'post_content' => wp_kses_post( wp_unslash( $raw['post_content'] ) ),
);
$result = wp_update_post( wp_slash( $data ), true );
if ( is_wp_error( $result ) ) {
    wp_die( 'Update failed.', '', array( 'response' => 500 ) );
}

Mass assignment writes client-selected fields that the endpoint never intended to expose. wp_update_post() does not perform the caller's capability check; its post_author and post_status fields must not become writable by accident. A user role is not a post field, but generic user/meta writers have analogous risks. Avoid extract( $_POST ) too: it lets request keys overwrite local variables.

For REST routes, args validates declared parameters; it does not automatically reject or remove undeclared parameters. Construct an explicit persistence payload rather than forwarding $request->get_params(). The Settings API can store one array option, but its sanitize_callback must validate shape and build an explicit allowlist of nested fields. Authorization and CSRF protection remain separate.

Related: see the output-escaping skill for the redirect "escape" context and the settings-options-security skill for recursively sanitizing nested option arrays.

Correct code examples

A complete reference of input types mapped to the correct sanitize/validate functions — text, email, int, float, key, slug, URL, HTML, arrays, JSON — is in references/sanitization-cheatsheet.md.

// Settings array sanitized field-by-field on save.
function my_plugin_sanitize_settings( $input ) {
    $clean = array();
    $clean['email']   = sanitize_email( $input['email'] ?? '' );
    $clean['count']   = absint( $input['count'] ?? 0 );
    $clean['enabled'] = ! empty( $input['enabled'] ) ? 1 : 0;
    $clean['note']    = wp_kses_post( $input['note'] ?? '' );

    if ( ! is_email( $clean['email'] ) ) {
        $clean['email'] = '';
        add_settings_error( 'my_plugin', 'bad_email', __( 'Invalid email.', 'my-plugin' ) );
    }
    return $clean;
}

Checklist

  • Every superglobal read is guarded with isset() and a default.
  • wp_unslash() is applied before sanitizing superglobal data.
  • Each value uses the sanitizer matching its type (see cheatsheet).
  • Rich HTML uses wp_kses / wp_kses_post, never strip_tags, for safety.
  • Values are validated against allowed sets / ranges after sanitizing.
  • Arrays are sanitized element-by-element (array_map).
  • URLs use esc_url_raw() for storage (esc_url() for display).
  • Bulk updates validate container/scalar shapes, authorize each target object, and build explicit field allowlists.
  • Settings API registrations supply a sanitize_callback.

Official references

Signals

GitHub stars
31
Forks
2
Last commit
Sep 2026
Advanced
Catalog kind
skill
Gateway key
input-sanitization-validation
Source
github.com/wpultimatesecurity/wordpress-security-skills