Agentic Product Lifecycle

mxa-audit

Provides the machinery the audit skills share: a JSON record that sits between what scripts measure and what a person judges, two entry points, and the rules that keep a score honest. It is not usually run on its own.

  • Writes one JSON record per audit that scripts fill and the agent reads, so measurement and judgement never blur.
  • Offers a full run, which measures, and a cheap re-render, which rewrites only the prose.
  • Leaves out anything it could not measure instead of scoring it, so a high mark means a check ran and passed.
  • Runs Lighthouse on every sampled page for the site audits, and stops with the reason when it cannot; there is no scanner of its own underneath.
  • Is read by the five audit skills; a person rarely invokes it directly.

What is inside

The scripts sample pages, drive Lighthouse, score and render, and write a JSON artifact two runs must agree on byte for byte. Nothing in them reads HTML or checks a page itself.

Uses
Lighthouseaxe-corePlaywright
Consulted
report-schema.json

Install it

npx skills add madebymany/madebyagents --skill mxa-audit
Read the skill itself