agentsclimarketplace

Magento audit

Skill staksoft/magento-claude-skills/skills/magento-audit

Performance audit of a Magento 2 / Mage-OS / Adobe Commerce storefront: full-page cache misses and cacheable="false" leaks, Varnish/Redis configuration, indexer modes and cron health, TTFB, payload sizes, and Core Web Vitals. Use this skill whenever the user says a Magento store is slow, asks to audit/review/check storefront performance, mentions FPC or cache hit rate, X-Magento-Cache-Debug, Varnish not caching, high TTFB, stuck or realtime indexers, cron problems, or Core Web Vitals / PageSpeed scores for a Magento or Mage-OS site. Works from a URL alone, from a codebase alone, or both. For writing or fixing module code, use the magento-module skill instead.From its SKILL.md

Install
npx -y skills add staksoft/magento-claude-skills --skill magento-audit

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

3 things to look at

  • reads credentialsReads from 1 credential source: `app/etc/env.php`.
  • 6 stars6 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.
  • runs commandsInstructs the agent to run 6 commands, including `python scripts/check-headers.py https://store.example/ https://store.example/some-category [--insecure]` and 5 more.

SKILL.md

5.3 KB, ~1.1k tokens by cl100k_base, as published. Nobody here has run it

Magento 2 / Mage-OS Storefront Performance Audit

Audit procedure for Magento storefront performance. The output is always a severity-ranked markdown report (references/scoring.md) where every finding carries verbatim evidence and an exact fix — never a list of generic tips.

The cardinal rule: audit caching first. Until full-page cache works, every other number (TTFB, CWV, server load) is measuring the wrong thing. Most "Magento is slow" reports are FPC failures with a different hat on.

Pick the mode

You haveModeTools
A store URLURL modescripts/check-headers.py, browser/Lighthouse if available
The codebase (and ideally a running install)Codebase modescripts/scan-layout.py, env.php, bin/magento CLI
BothCombined — do URL mode first, use codebase mode to explain what it foundboth

Ask for the missing half only if the findings demand it (e.g. URL mode shows MISSes → request codebase access to find the killer); otherwise audit what you have and record the gap in the report's "Not audited" section.

URL mode

  1. Collect 3–5 public URLs: homepage, a category page, a product page, a CMS page. Never test cart/checkout/account pages for cache hits — they are uncacheable by design.

  2. Run the header check (stdlib-only Python, two requests per URL so the second is warm):

    python scripts/check-headers.py https://store.example/ https://store.example/some-category \
        [--insecure]   # for local/dev self-signed certs
    

    It measures TTFB cold/warm, payload size, compression, and parses X-Magento-Cache-Debug / Age / Varnish headers into pre-classified findings.

  3. Interpret with references/fpc-audit.md — particularly the header table and what a warm MISS means.

  4. Frontend layer: read references/frontend-perf.md; run Lighthouse/PageSpeed if available, otherwise estimate from the fetched HTML (script/CSS counts, LCP image preload, lazy-loading mistakes, third-party tags).

Codebase mode

  1. FPC killers — scan layout XML for cacheable="false":

    python scripts/scan-layout.py /path/to/magento [--include-vendor] [--json]
    

    Severity is pre-classified (default.xml → critical, catalog/CMS handles → critical, checkout/customer → info). For runtime killers the scan can't see (isScopePrivate, session starts in frontend blocks), follow references/fpc-audit.md.

  2. env.php and stack — read app/etc/env.php against references/server-config.md: cache/session backends and Redis DB separation, http_cache_hosts, deploy mode, OPcache. Never quote secrets (passwords, keys) from env.php into the report — name the key, not the value.

  3. Indexers & cron — with a running install:

    bin/magento indexer:show-mode      # any 'realtime' on production = finding
    bin/magento indexer:status
    bin/magento cache:status           # full_page enabled?
    bin/magento deploy:mode:show
    

    plus the cron_schedule checks in references/indexers-cron.md. Without a running install, audit crontab.xml groups and env.php cron_consumers_runner statically and note the rest as not-audited.

Write the report

Assemble findings using the template and scoring rubric in references/scoring.md. Non-negotiables:

  • Every finding: verbatim evidence (header value, file:line, CLI output) + one concrete fix + effort estimate.
  • Rank by impact-per-effort. A critical FPC leak outranks ten image optimizations.
  • Include the "Working as intended" section — name the scary-looking-but-correct things (uncached checkout, cold-request slowness) so nobody "fixes" them.
  • Include "Not audited" — what this mode couldn't see.
  • For dev/staging targets, downgrade production-only findings (developer mode, exposed debug headers) to info and say absolute timings need re-measuring on production.

Audit, don't fix

The deliverable is the report. Don't start editing layout XML or env.php unless the user asks for fixes — then switch to the magento-module skill conventions for any code changes, and re-run the relevant script afterwards to prove the finding is gone.

Pairing with live data

If the elgentos magento2-dev-mcp MCP server is connected, prefer it for reading merged layout/config when chasing an FPC killer — Magento merges XML across modules and themes, and the single-file view can mislead.

What ships with it: 7 files

41.5 KB alongside SKILL.md, 2 of them executable

scripts/

Gives 0 of the 12 instructions most audit compliance skills give in ~1.1k tokens

Counted across 960 of the 1,589 authors here whose files we hold, read 2026-09-06

  • Read product marketing context before asking questionsin 29 of 960, across 11 files
  • Rank findings by severityin 29 of 960, across 22 files
  • Generate audit reportin 22 of 960
  • Run the audit scriptin 20 of 960, across 19 files
  • Generate a prioritized action plan reportin 19 of 960, across 11 files
  • Ensure one H1 per pagein 15 of 960, across 5 files
  • Ensure sitemap exists and is accessiblein 14 of 960, across 4 files
  • Verify alt text on all imagesin 12 of 960, across 3 files
  • Determine the audit scope before startingin 12 of 960, across 4 files
  • Verify important pages allowed in robots.txtin 11 of 960, across 2 files
  • Detect business type from homepage signalsin 11 of 960, across 7 files
  • Delegate specialized tasks to subagentsin 11 of 960, across 7 files

Said here and by no other author read

  • Audit caching first
  • Collect public URLs for testing
  • Run the header check script
  • Scan layout XML for cacheable false
  • Read app etc env php
  • Check indexer and cron status

Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once. Length counted with cl100k_base; the agent that loads this file may tokenize it differently.

Keep looking

Skills are one crate of 325,949. 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.