agentsclimarketplace

Wp plugin rewrite rules

Skill Lonsdale201/wp-agent-skills/plugin-scaffold/wp-plugin-rewrite-rules

A community-maintained collection of agent skills for WordPress plugin and theme development.

Install
npx -y skills add Lonsdale201/wp-agent-skills --skill wp-plugin-rewrite-rules

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 21 stars21 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Design and review custom WordPress URL rewrites: add_rewrite_rule, add_rewrite_tag, query_vars, CPT/taxonomy rewrite slugs, add_rewrite_endpoint, soft vs hard flushes, rewrite_rules cache behavior, and the rule that flush_rewrite_rules() must not run on every request. Use for custom pretty URLs, CPT permalink 404s, endpoint rewrites, and code containing flush_rewrite_rules or add_rewrite_rule.

SKILL.md

16.4 KB, as published. Nobody here has run it

WordPress plugin: rewrite rules & flush

The single most consistently-mishandled topic in WP plugin code by AI assistants and inexperienced developers. The pattern that "works" — call flush_rewrite_rules() after add_rewrite_rule() — is correct but the WHEN matters more than the WHAT. Calling it once is right; calling it on every page load wrecks performance.

This skill covers when and how to register custom URL rewrites and the discipline around flushing.

When to use this skill

Trigger when ANY of the following is true:

  • Adding a custom URL endpoint (/api/v1/..., /dashboard/orders/, /track/<token>/).
  • Registering a CPT or taxonomy with a 'rewrite' argument that creates new permalink structures.
  • Adding a "REST-but-pretty-URL" pattern that uses WP's rewrite engine instead of register_rest_route.
  • Debugging "my new permalink returns 404 even though the post exists".
  • Reviewing code where you see flush_rewrite_rules() or add_rewrite_rule() — verify the placement.

Mental model — rewrite_rules is a cached option

WordPress's permalink engine works in two passes:

  1. Generate: based on registered CPTs, taxonomies, custom rules (add_rewrite_rule), endpoints (add_rewrite_endpoint), and the configured permalink structure, WP builds a big regex array. Each entry maps a URL pattern → query variables.
  2. Cache: this entire array is stored in the rewrite_rules option (wp-includes/class-wp-rewrite.php refresh_rewrite_rules(), since WP 6.4). Whether the option is autoloaded depends on its stored autoload value and newer WP autoload heuristics; do not assume every install has the same value.
  3. Match: every request uses the cached option to figure out what query to run.

Generation is expensive (iterates ALL post types, taxonomies, endpoints, custom rules). The cache is regenerated only when explicitly told to — via flush_rewrite_rules() or by the user saving the Permalinks settings page.

The footgun: register a CPT, see 404 on the CPT permalink, and "fix it" by adding flush_rewrite_rules() to the init callback. This works (the second visit succeeds), but you've now made every page request rebuild + write the rewrite_rules option AND rewrite .htaccess. Sites with many CPTs see noticeable slowdowns.

The flush rule — once on activation, once on deactivation, NEVER on init

// === Runtime: registers the rewrite on EVERY request, no flush ===
add_action( 'init', static function (): void {
    add_rewrite_rule(
        '^api/v1/items/?$',
        'index.php?myplugin_endpoint=items',
        'top'
    );
} );

add_filter( 'query_vars', static function ( array $vars ): array {
    $vars[] = 'myplugin_endpoint';
    return $vars;
} );

// === Activation: flush ONCE, after CPTs/rewrites are registered ===
register_activation_hook( __FILE__, static function (): void {
    // Make sure our rewrite is registered for THIS request before flushing,
    // since activation runs after `init` of the activation request.
    add_rewrite_rule(
        '^api/v1/items/?$',
        'index.php?myplugin_endpoint=items',
        'top'
    );
    flush_rewrite_rules();
} );

// === Deactivation: flush ONCE, so the generated option can be rebuilt ===
register_deactivation_hook( __FILE__, static function (): void {
    flush_rewrite_rules();
} );

Three rules:

  1. Register rewrites on init, every request. Cached rules can keep matching until the next flush, but registration must be present whenever WP regenerates the rules; otherwise the next flush drops your rule.
  2. Flush on activation and deactivation only. That's two writes per plugin lifecycle, not two per request.
  3. Don't conditionally flush at runtime ("if rules don't include mine, flush"). It's tempting; it's also fragile and easy to mis-trigger.

If a settings UI change affects URL structure (a slug rename), flush after the option is saved — gated by the specific setting having changed:

add_action( 'update_option_myplugin_settings', static function ( $old, $new ): void {
    if ( ( $old['slug'] ?? '' ) !== ( $new['slug'] ?? '' ) ) {
        flush_rewrite_rules();
    }
}, 10, 2 );

That's the only place runtime flushing belongs. A settings screen flush should almost always use flush_rewrite_rules( false ) unless the permalink file rules also changed.

CPT and taxonomy rewrite slugs

register_post_type( 'foo', array( 'rewrite' => array( 'slug' => 'foos' ) ) ) adds permalink rules for foos/<post-slug>/. WP does not auto-flush when you register the CPT (verified in wp-includes/post.php — no flush_rewrite_rules() call inside register_post_type). Same for register_taxonomy.

This is why CPT-based plugins MUST flush on activation. Otherwise: install the plugin → first CPT entry created → 404 on the entry's permalink → user thinks plugin is broken.

register_activation_hook( __FILE__, static function (): void {
    myplugin_register_post_types();   // call your CPT registration
    flush_rewrite_rules();
} );

If your CPT registration is encapsulated inside a class method that's normally only called on init, expose it as a callable function or static method so the activation hook can call it directly. Don't rely on the init of the current request having already registered it (in some flows, like programmatic activation via activate_plugin(), hook order is shifted).

Custom rewrite rules — the add_rewrite_rule + handler pattern

To handle a URL pattern that doesn't map to a post or taxonomy:

// 1. Register the rule on init.
add_action( 'init', static function (): void {
    add_rewrite_rule(
        '^track/([a-z0-9]+)/?$',          // URL pattern
        'index.php?myplugin_track=$matches[1]',  // query string
        'top'                             // priority over default rules
    );
} );

// 2. Whitelist the query var (otherwise WP strips it).
add_filter( 'query_vars', static function ( array $vars ): array {
    $vars[] = 'myplugin_track';
    return $vars;
} );

// 3. Handle the request — usually on template_redirect or parse_request.
add_action( 'template_redirect', static function (): void {
    $token = get_query_var( 'myplugin_track' );
    if ( ! $token ) {
        return;
    }
    // Render or redirect. exit; if you don't want WP's template loader to also run.
    myplugin_render_track_page( sanitize_key( $token ) );
    exit;
} );

// 4. Activation: flush after registering.
register_activation_hook( __FILE__, static function (): void {
    add_rewrite_rule( '^track/([a-z0-9]+)/?$', 'index.php?myplugin_track=$matches[1]', 'top' );
    flush_rewrite_rules();
} );

Notes on each step:

  • add_rewrite_rule( $regex, $redirect, $position ) (wp-includes/rewrite.php). $position: 'top' matches before WP defaults (use for plugin endpoints that should override ?p=NN style queries); 'bottom' matches after (use for fallback patterns).
  • The query var IS whitelist-required. Without the query_vars filter entry, get_query_var() returns empty even if the rule matched. If you use add_rewrite_tag( '%myplugin_track%', '([a-z0-9]+)' ) before building rules, WordPress registers the matching public query var for you.
  • template_redirect is the usual handler hook for rendering custom output from a rewrite. parse_request runs earlier if you need to short-circuit before WP_Query.
  • REST is usually a better choice for true API endpoints — register_rest_route gives you typed args, permission_callback, JSON formatting, and cookie auth for free. Use add_rewrite_rule only when you genuinely need pretty URLs that participate in WP's permalink engine (frontend pages, redirects, content-driven routes).

add_rewrite_endpoint — extending existing permastructs

For URLs that hang off existing permalinks, like /<post-slug>/json/ or /blog/<post-slug>/print/:

add_action( 'init', static function (): void {
    // EP_PERMALINK | EP_PAGES — append /json/ to single-post and page permalinks
    add_rewrite_endpoint( 'json', EP_PERMALINK | EP_PAGES );
} );

// On the template hook:
add_action( 'template_redirect', static function (): void {
    if ( get_query_var( 'json' ) !== '' && is_singular() ) {
        wp_send_json( myplugin_serialize_post( get_post() ) );
        exit;
    }
} );

add_rewrite_endpoint automatically registers a query var matching the name by default and adds rules to the relevant permastructs. Pass false as the third argument to skip query-var registration, or a string to use a custom query-var name. Bitmask options (verified in wp-includes/rewrite.php and EP_* constants in wp-includes/class-wp-rewrite.php):

  • EP_PERMALINK — single posts
  • EP_PAGES — pages
  • EP_ALL — every permastruct WP knows
  • EP_ROOT — root only (/json/)
  • EP_CATEGORIES, EP_TAGS, EP_AUTHORS, etc.

Same flush rule applies: register on init, flush on activation.

Soft vs hard flush

flush_rewrite_rules( $hard = true ) (wp-includes/rewrite.php):

  • Hard flush ($hard = true, default): rebuilds the rules array, writes the rewrite_rules option, AND writes .htaccess (Apache) / web.config (IIS) with the regenerated mod_rewrite directives.
  • Soft flush ($hard = false): rebuilds rules array, writes the option only. Skips the file write.

For activation / deactivation: hard flush is correct (the file edit catches setups where mod_rewrite reads from .htaccess directly). For settings-change-driven flushes mid-runtime: soft is enough if .htaccess is unchanged.

The .htaccess write requires file-system write access in the WP install directory. Hosts with read-only deployment may emit a notice or silently skip it; the soft flush still updates the option, which is what most lookups use anyway.

The rewrite_rules option — cached and growable

rewrite_rules is stored in wp_options and is read whenever WP needs to match pretty permalinks. Many installs have it autoloaded, but WP 6.6+ may store automatic autoload decisions such as auto, auto-on, or auto-off; audit the actual value before making a performance claim. A site with:

  • 50 plugins each adding 2-3 custom rules
  • 20 CPTs
  • 5 custom endpoints

…can end up with a rewrite_rules option in the 100KB-500KB range. If autoloaded, that's read into memory on every uncached request. Two consequences:

  • Don't add rewrites you don't need. Each rule lives in the option forever (until flushed away by deactivation).
  • Plugin uninstall should ensure cleanup. Deactivation usually flushes, but the user might delete the plugin without deactivating. Add a flush_rewrite_rules() call to uninstall.php if your plugin added rules — see wp-plugin-lifecycle.

You can audit the option's size with:

SELECT LENGTH(option_value) AS bytes, autoload
FROM wp_options
WHERE option_name = 'rewrite_rules';

A healthy site is in the 10-50KB range. 200KB+ is a smell.

Multisite

Each blog has its own rewrite_rules option — flush is per-blog. For multisite-network plugins that add rewrites, flush per-site at activation:

register_activation_hook( __FILE__, static function ( bool $network_wide = false ): void {
    if ( $network_wide ) {
        foreach ( get_sites( array( 'fields' => 'ids' ) ) as $site_id ) {
            switch_to_blog( $site_id );
            myplugin_register_rewrites();
            flush_rewrite_rules();
            restore_current_blog();
        }
    } else {
        myplugin_register_rewrites();
        flush_rewrite_rules();
    }
} );

Multisite caveat: this skill's authoring environment is single-site. The pattern above is source-derived (switch_to_blog (wp-includes/ms-blogs.php) is per-blog, flush_rewrite_rules operates on the current blog). Verify on a real network install.

Critical rules

  • Register rewrites on init (every request); they have to exist when the rules are regenerated.
  • Flush on activation + deactivation, NEVER on init. The single highest-leverage rule in this skill.
  • register_post_type / register_taxonomy do NOT auto-flush — your activation handler must.
  • Whitelist custom query vars via the query_vars filter, or get_query_var() returns empty.
  • Prefer REST (register_rest_route) for API endpoints; use rewrite rules for permalink-engine-participating URLs.
  • Match settings-change flushing to specific setting changes, not blanket "after any save".
  • Hard flush = file write; soft flush = option write only. Default $hard = true is fine for activation.
  • Audit rewrite_rules option size if you suspect rule bloat. 200KB+ is a smell.

Common mistakes

// WRONG — flush on every request, .htaccess rewritten constantly
add_action( 'init', function () {
    add_rewrite_rule( '^api/v1/items/?$', 'index.php?myplugin_endpoint=items', 'top' );
    flush_rewrite_rules(); // disaster
} );

// WRONG — registers a CPT but never flushes; permalinks 404
register_activation_hook( __FILE__, function () {
    register_post_type( 'myplugin_log', array( 'rewrite' => array( 'slug' => 'logs' ) ) );
    // missing: flush_rewrite_rules();
} );

// WRONG — flush on activation but rule isn't registered for this request
register_activation_hook( __FILE__, function () {
    flush_rewrite_rules(); // rebuilds rules WITHOUT my CPT (registered on init only)
} );

// WRONG — query_var not whitelisted; get_query_var returns empty
add_action( 'init', function () {
    add_rewrite_rule( '^track/([a-z0-9]+)/?$', 'index.php?myplugin_track=$matches[1]', 'top' );
} );
// Missing: add_filter( 'query_vars', fn ( $v ) => array_merge( $v, array( 'myplugin_track' ) ) );

// WRONG — using add_rewrite_rule for a JSON API
add_rewrite_rule( '^api/v1/orders/?$', 'index.php?myplugin_orders=1', 'top' );
// Better: register_rest_route( 'myplugin/v1', '/orders', ... );

// WRONG — flushing on every settings save regardless of what changed
add_action( 'update_option_myplugin_settings', 'flush_rewrite_rules' );
// Should be gated to specific URL-affecting fields changing.

Cross-references

  • Run wp-plugin-lifecycle for the activation/deactivation/uninstall placement of flush_rewrite_rules() in full lifecycle context.
  • Run wp-rest-api when the endpoint is a JSON API — REST is almost always a better choice than rewrite-rule + custom handler.
  • Run wp-plugin-options-storage if rewrite_rules option size is the concern; the skill explains autoload tradeoffs in general.

What this skill does NOT cover

  • The WP_Rewrite class internals beyond what flush_rules does. Most plugin code shouldn't touch the class directly.
  • Multisite-network domain mapping plugins (Multisite-style URL routing across subsites) — niche, separate concern.
  • mod_rewrite vs nginx rewrite block differences. WP writes mod_rewrite to .htaccess for Apache; nginx requires manual location directives in the server config (no automatic write from WP).
  • URL canonicalization (redirect_canonical) and trailing-slash handling — adjacent topic.
  • Page-builder / theme-introduced rewrite gotchas.

References

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.