Audit UX and Save Files

Disciplined usability evaluation of a running app, from the perspective of the people who actually use it — not a frontend pass. The deliverable is a durable on-disk artifact of usability problems and a prioritized list of the major changes they imply.

Dispatch parallel audit subagents that each embody one persona plus a UX expert's eye, drive the run-agent-browser skill through that persona's core journeys, capture screenshots at the moments friction appears, and write one markdown finding per real usability problem into a dated, persona-scoped, journey-scoped tree under ux-findings/. When the audit returns, cluster findings into UX themes, then synthesize a prioritized recommendations report. This skill reports; it does not fix.

This is not "make it look nicer." The questions are: Can this persona accomplish their goal? Where do they hesitate, misread, backtrack, or give up? What is confusing, redundant, or unnecessary? What one change would help most? Findings are judged by task impact and business consequence, not by pixels.

The skill owns three things: where files live, how usability findings are written (persona · journey · heuristic · severity · behavioral + business impact · recommended major change), and how the audit is synthesized into recommendations. It does NOT own visual/CSS correctness (that is audit-ui-and-save-files) and does NOT implement the recommendations.

When to Use

Trigger on phrases and contexts like: