v0.6.0 45: Run release usability study

[WilsonLe]

Epic

Epic G — Hardening and release

Goal

Run release usability study for the v0.6.0 minimal task-first UI/UX release.

Acceptance criteria

• Run final moderated task study for first transcript, vocabulary, enrollment, training, export, import, delete-data tasks.
• Check 90% unassisted completion gates and comprehension gates from the plan.
• Fix or document high-severity findings through a release decision.
• Do not claim gates passed without evidence.

Common v0.6.0 constraints

• Preserve all v0.5.0 speech, enrollment, browser-training, profile, portability, privacy, security, and compatibility contracts unless a separate issue proves a contract need.
• Keep heavy work in existing packages/workers; do not move inference, training, archive parsing, encryption, hashing, or large storage work onto the UI thread.
• Keep local-first privacy: no telemetry, no new remote service, and no personal audio/transcript/vocabulary/model/profile data leaving the browser.
• Default UI must reduce copy and technical detail while keeping required actions, blockers, consequences, and recovery visible.
• Maintain WCAG 2.2 AA target, keyboard operation, screen-reader semantics, responsive layout, reduced-motion, forced-colours, and zoom/reflow behaviour.

Dependency order

This is item 45 of 48 in the v0.6.0 dependency-ordered backlog.

[WilsonLe]

Reopening: PR #303 records only the release-usability insufficient-evidence blocker. It does not complete the final moderated study or satisfy the release usability gate. Keep this issue open until consented participant evidence is added or an explicit human release decision is recorded.