performance
Analyzes WordPress performance across database queries, PHP execution, frontend rendering, and caching layers. Use when investigating slow pages, optimizing load times, or reviewing caching strategy.
No fetch-and-execute, destructive, credential, exfiltration or rules-override patterns in 1 files, checked on Oct 20, 2026. This is not a guarantee of safety. How the scan works.
npx degit Shaisolaris/solaris-dev-shop/employees/specialized/wordpress-master/diagnostics/performance .claude/skills/performance12 alternatives · Best frontend skills · Author? Prove this skill
SKILL.md
# Domain 3: Performance Analysis ## 3.1 Database Performance - Identify N+1 query patterns (queries inside loops) - Flag `meta_query` without proper indexing - Detect unnecessary `LIKE '%value%'` queries (leading wildcard prevents index use) - Check for missing indexes on custom tables - Identify excessive `wp_options` autoloading (total autoloaded data size) - Review transient usage - orphaned or never-expiring transients - Detect `SELECT *` where specific columns would suffice - Flag `ORDER BY RAND()` on large datasets - Check post revision accumulation and recommend limits - Review `wp_postmeta` table bloat ## 3.2 PHP Performance - Identify expensive operations in `init`, `wp_loaded`, or `admin_init` hooks that should run conditionally - Detect heavy computation on every page load vs lazy/on-demand execution - Flag blocking HTTP API calls without timeouts - Identify missing object caching for repeated expensive operations - Check for excessive use of `get_option()` on non-autoloaded options (each triggers a DB query) - Detect memory-intensive operations (loading all posts, large arrays in memory) - Review cron jobs for efficiency and proper scheduling - Identify duplicate plugin functionality (multiple SEO plugins, multiple caching layers) ## 3.3 Frontend Performance - Audit render-blocking CSS and JavaScript - Check proper use of `wp_enqueue_script()` / `wp_enqueue_style()` (not hardcoded in templates) - Verify scripts load in footer where possible (`in_footer` parameter) - Identify scripts/styles loaded globally that should be conditional - Check for proper async/defer attributes - Review image optimization (WebP support, proper sizing, lazy loading) - Identify excessive DOM size from page builder bloat - Check for unused CSS/JS being loaded - Verify proper use of WordPress asset dependency system - Core Web Vitals impact assessment (LCP, FID/INP, CLS) ## 3.4 Caching Analysis - Page caching configuration and effectiveness - Object caching (Redis/Memcached) configuration - Browser caching headers (Cache-Control, Expires, ETag) - CDN configuration and cache hit rates - Fragment caching for dynamic content areas - Transient API usage patterns - OPcache configuration for PHP --- ## Tool Usage: Performance Analysis Commands ### Database Analysis ```bash # Analyze autoloaded options size wp db query "SELECT option_name, LENGTH(option_value) as size FROM wp_options WHERE autoload='yes' ORDER BY size DESC LIMIT 20;" # Count orphaned transients wp db query "SELECT COUNT(*) FROM wp_postmeta WHERE meta_key LIKE '_transient_%';" # Count post revisions wp db query "SELECT COUNT(*) FROM wp_posts WHERE post_type='revision';" # Analyze wp_postmeta table size wp db query "SELECT COUNT(*) as total_rows, SUM(LENGTH(meta_value)) as total_size FROM wp_postmeta;" # Find slow queries (requires query log enabled) wp db query "SELECT * FROM mysql.slow_log ORDER BY query_time DESC LIMIT 10;" ``` ### WordPress CLI Diagnostics ```bash # Database optimization wp db optimize wp db repair # Transient cleanup wp transient delete --expired wp transient delete --all # Cron inspection wp cron event list wp cron schedule list # Cache management wp cache flush wp cache type # Show cache type (Redis, Memcached, etc.) ``` ### Query Monitoring ```bash # Enable Query Monitor plugin for development wp plugin install query-monitor --activate # Check for slow queries in debug log tail -f /path/to/wp-content/debug.log | grep "Slow query" ```
Shown under the source repo’s MIT license. Read the full folder on GitHub.