wp-performance_skill

Overview

This skill helps investigate and improve WordPress backend performance without a browser UI. It focuses on profiling, deterministic measurements, database and autoload optimization, object caching, cron behavior, and safe verification workflows. Use it to generate reproducible backend performance reports and guide targeted fixes.

How this skill works

The skill runs automated backend inspections using WP-CLI and HTTP requests to capture baseline TTFB and profile data. It detects WP-CLI/profiler availability, autoloaded option sizes, object-cache drop-ins, and emits recommendations for fast wins or deeper profiling. For headless Query Monitor usage it reads REST headers or envelope responses to surface slow queries and hooks.

When to use it

• A WordPress endpoint, admin screen, or REST route has high TTFB or sporadic slowness
• You need a safe, backend-only profiling plan (no browser required)
• You want to find and fix DB query hotspots, autoload bloat, or cache misses
• Investigating WP-Cron spikes or expensive remote HTTP calls
• You must profile in environments with restrictions on installs or UI access

Best practices

• Measure first: capture baseline TTFB and WP-CLI profile output before changes
• Confirm environment and write permissions (dev/staging vs prod) and any operation restrictions
• Prefer deterministic WP-CLI tools: wp doctor, wp profile stage/hook/eval
• Target one dominant bottleneck category and fix iteratively, then re-measure
• Avoid risky ops in production: no unsolicited plugin installs, SAVEQUERIES, or cache flushes during traffic

Example use cases

• Generate a backend performance report for a slow REST route using node skills/wp-performance/scripts/perf_inspect.mjs
• Run wp doctor check to detect autoloaded option bloat before deep profiling
• Use wp profile hook with --url to identify a slow plugin callback on a specific endpoint
• Inspect Query Monitor REST headers (x-qm-*) or ?_envelope responses to find expensive queries headlessly
• Reduce cron-induced spikes by de-duplicating events and moving heavy tasks off request paths