ha-panel-ux-auditlisted

Runs a read-only panel-level UX audit of an existing panel artifact (custom sidebar panel, panel-mode view, or custom view) — with mobile-device usability as a mandatory first-class dimension. Audits against spec/ha/lovelace-views-panels, spec/ha/lovelace-layout-antipatterns, spec/ha/lovelace-card-patterns, and spec/ha/frontend-data-api plus UX/accessibility heuristics, and produces a severity-sorted improvement report — mobile usability, responsive/narrow behaviour, touch targets, information hierarchy, readability, accessibility, perceived performance — that ha-panel-author can consume as a prioritized work-list. Never modifies code. Activate on "audit my panel's UX", "is this panel usable on mobile", "review the panel for usability and accessibility", or equivalent German requests. Do not activate for building or fixing a panel (ha-panel-author), scaffolding one (ha-panel-add), a whole-integration review (ha-integration-reviewer), or deploying to a live HA instance.
nolte/claude-home-assistant · ★ 1 · Web & Frontend · score 68

Install: claude install-skill nolte/claude-home-assistant

# HA Panel UX Audit Spec: `spec/claude/ha-panel-ux-audit/en.md` (EN canonical) / `spec/claude/ha-panel-ux-audit/de.md` (DE translation). This skill is the **UX expert** of the panel family. It statically audits an existing panel artifact from the user's point of view — can a person actually use this, especially on a phone — and emits a severity-sorted report of concrete improvements. It never edits code; the report is meant to be read, acted on, and — when the user wants — handed to `ha-panel-author` as a prioritized work-list. ## Why this is a skill, not an agent - **Human-visible audit surface** — like the sibling audit skills `ha-quality-scale-audit` and `ha-security-audit`, this is an interactive audit the user invokes directly and reads the report from; a skill keeps it on the visible command surface. - **Mid-flow interactivity** — the target devices (mobile is mandatory, but which else), the audience, and severity threshold are per-run inputs the user confirms before the audit. - **Orchestrator-leaning** — findings route to `ha-panel-author` for the fix; the skill-orchestrates-fixes default keeps the entry point in skill form. - Counter-dimension considered: a read-only one-shot review could be an agent (cf. `ha-integration-reviewer`), but the report is meant to be read and acted on interactively and to feed `ha-panel-author`, and consistency with the `ha-*-audit` skills wins. ## When this skill activates Use this skill to audit an existing **panel-level** artifac