ce-ux-auditlisted

# Ux Audit **Invocation input:** Journey, scope, or running app to audit (optional): $ARGUMENTS Walk a running app's UX via the browser and report **evidence-backed findings** an agent can reliably detect — dead ends, plan-vs-reality gaps, cross-feature inconsistencies, broken links, missing states, validation gaps, state loss, layout breakage, accessibility violations. The workflow is **read-only on code and existing artifacts**. It finds issues and **escalates** to the rest of the toolset (`/ce-implement`, `/ce-spec`, `/ce-plan`) — it never patches code or modifies any other artifact. Same discipline as `/ce-verify`. It runs in one of **two auto-detected modes**, chosen by a Stage-0 probe of plan state — you never have to know which: - **Journey-walk mode** — a plan on disk owns the surface. *Verifies* each planned journey against its EARS criteria: walk it once, run the mechanical checks, report where the running app diverges from the traced plan. - **Adversarial-discovery mode** — no plan owns the target (a plain running app, or a repo with no traced journeys). *Discovers* unknown problems by exploring journeys the planner never traced and probing inputs and states no spec specified — a chaos taxonomy over inputs, sequence, auth, viewport, state. Same evidence-bound finding shape and the same triage discipline in both modes; different inputs and different jobs (verification vs discovery). ## Runtime Inputs - **Target (mode-specific):** in **journey-walk