Agentic Product Lifecycle

mxa-verify-promote

Stands in front of production: a full regression, performance measured on real content, the manual and assistive-technology passes a machine cannot do, and a recorded sign-off. It is where user acceptance testing lands.

  • Runs the whole journey suite against a production-like build with real content, not fixtures.
  • Re-runs the performance budget on the real thing, where fixture numbers usually flatter.
  • Schedules the keyboard, screen-reader and real-device passes, with a named person for each.
  • Records the sign-off against the specific deployment, with the dated audit reports it relied on.
  • Prepares everything and promotes nothing, because going live is a person's decision.

What is inside

Instructions. Nothing here runs on its own: the judgement is the skill.

Evals
3

Works with

The outside tools and connectors this skill drives.

  • Playwright regression tests· our default
  • Lighthouse on real content
  • recorded human checks

The first is what we reach for on our own projects. A project that already uses one of the others keeps it, and the skill works with that instead.

Install it

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