Validating Changes in the Browser

This skill is the last station before a ticket is shipped. It is invoked only after the implementation is complete and the relevant review step has succeeded — not as a substitute for either, but as the final user-eye check that catches the things a unit test, an API test, or a static review cannot see: broken empty states, missing toasts, focus traps, regressed flows on roles the developer didn't think to test, console errors, layout glitches, latent bugs in sibling pages that the change accidentally exposed.

The orchestrator runs the test list itself via Chrome DevTools MCP. Whatever it finds — including issues that pre-date the current branch — is fixed in the same loop. The user only sees the final list, fully ticked, with a question: ship or optimize further.

Language convention for this skill:

• The instructions you read here (English) tell you HOW to run the workflow.
• The artefacts the developer sees at the end (test plan headings, walked list, status labels, ship-or-optimize question and options) are rendered in the language the user has been speaking in this session — German if the conversation has been in German, English if in English, the user's language otherwise. The English examples below are illustrative templates; translate the headings, columns, action prose, status labels, and the AskUserQuestion text+options to match the user's language before showing them.

When to Use This Skill

Activates at the end of these workflows (invoked from each):