/goal Turn the existing participant and Rita screens into a calm, understandable mobile practice experience: the learner knows what to do today, can submit a video, and sees relevant personal feedback before continuing.

# Contract 04-course, revision 1: A daily practice app and a usable Rita feedback loop

## Goal
Turn the existing participant and Rita screens into a calm, understandable mobile practice experience: the learner knows what to do today, can submit a video, and sees relevant personal feedback before continuing.

## Scope and authority
Read `/Users/agency/Documents/Agty/KomplexLogopédia/plans/rita-ui-parallel-20261002/SHARED.md` first. Its product facts, ownership, routes, verification rules and completion boundaries are part of this goal. This is authorised frontend improvement when Matt launches this contract, not permission to launch the other contracts.

Work only under `/Users/agency/Documents/Agty/KomplexLogopédia/outputs/rita-ui-parallel-20261002/site`. Your exclusive UI ownership: **app/** · assets/daily/****. You also own `/Users/agency/Documents/Agty/KomplexLogopédia/outputs/rita-ui-parallel-20261002/handoffs/04-course/`, `/Users/agency/Documents/Agty/KomplexLogopédia/outputs/rita-ui-parallel-20261002/evidence/04-course/`, and `/Users/agency/Documents/Agty/KomplexLogopédia/outputs/rita-ui-parallel-20261002/reports/04-course.md`. Other agents share the workspace; preserve their edits. The old executor and the immutable baseline are read-only.

## Authoritative inputs and fixed decisions
Start with the existing app shell, app.js/app.css, fourteen lessons, fourteen expert screens, six original source texts and 56 existing visual alternatives. Both sides of the feedback loop stay with one owner because they share code and participant/day state.

The full existing site is already copied to the new work root. Do not rebuild from zero. The original snapshot and its hashes are at `/Users/agency/Documents/Agty/KomplexLogopédia/plans/rita-ui-parallel-20261002/baseline-manifest.json`. Follow the shared source hierarchy, brand, daily-feedback model and frontend-only boundary.

## Required deliverables
Refine entry/login demonstration, course overview, days01–14, expert days01–14 and the final consultation interface. Keep two infographic and two AI-image choices per day, 56 total, with a calm recommended choice in the customer flow and the alternatives in review controls. Preserve all full lesson content and the current local video/state mechanisms where they already work.

The acceptance owner is the canvas integrator for module handoff and Matt for the eventual client presentation. Final planned canvas URL: https://review.clientsflow.hu/komplex-rita-ui-canvas-2026-10-02-v1/ . This is a destination to create, not a claim that it is already published.

## End-state checklist
### APP1. Every day is a usable lesson, not an archive page

Criteria:
- All 14 day URLs show the right day/title, what to do first, relevant exercise instructions, what to record, reflection/sliders, submission state and next action. Rita's original exercise wording and source cautions remain complete and reachable; no duplicate contradictory simplified instruction is substituted.
- An independent reviewer can perform the day from the screen alone without guessing the exercise order or what video to send. Use progressive disclosure for long originals and optional visuals while keeping today's instructions visible. Verify all six originals against their stored hashes and map each of the 14 days to the intended source.
- Days2/5/8/12 begin with the prior relevant video/feedback when available; missing submission, waiting feedback, received feedback and missed-day/return states remain understandable. Day1's spontaneous clip and day14's agreed 2–3-minute practice clip are not confused.

### APP2. The participant-to-Rita-to-participant loop works locally

Criteria:
- For every one of 14 day IDs: change a slider and reflection → save → choose/preview a valid video → submit → open the matching Rita screen → read that day's task and participant input → draft/preview/send a written response → read that same response in the participant screen. Required state persists through refresh within this browser demo.
- Rita sees participant identity, day, submission status, the playable correct clip, task context and previous useful feedback without navigating a maze. Empty/missing/invalid video and unsent drafts never masquerade as reviewed work. The participant sees waiting versus received feedback distinctly.
- Check both existing sample participants to catch state mixing. Changing day or participant must never show another person's clip, reflection or feedback. Keep technical fixtures marked as samples, not Rita teaching videos or genuine personal assessments.

### APP3. Completion and consultation feel like a natural continuation

Criteria:
- The overview makes current, completed and upcoming work easy to recognise without locking a reviewer out. On mobile at360/390/430, video controls, sliders, long Hungarian instructions, keyboard/form state and primary buttons remain readable and usable.
- After day14 the learner can review what they practised, provide the final clip, choose a clearly sample consultation time, see confirmation/preparation, reschedule/cancel, and inspect a post-consultation next-step view. No fabricated real availability, real Meet link or claim that all development is complete.
- All 56 existing visuals remain reachable by day with two infographic/two image alternatives. Their content matches the actual exercise and does not introduce an unsafe or contradictory instruction. Luna inspects actual daily content and interface states; broken or misleading visuals are repaired selectively.

## Verification and evidence
For every day capture the opening/task section, usable exercise content, reflection/video state and the matching Rita screen. Luna must confirm the learner knows the next action and Rita has the correct context. Also capture saved→submitted→waiting→received feedback with identical day/participant identifiers, the four rewatch checkpoints, both participant identities, final clip and consultation states. Mobile only. The reviewer must read the actual lesson text, not just check a heading or green status.

Keep screenshots and findings in your assigned evidence directory. Link actual observations to the criterion IDs above. A screenshot placeholder is a pending check, never proof. Put useful UI first and evidence behind its concise handoff. Before declaring a route ready, use it in the rendered browser and read its entire unique copy. Use the required independent Luna process from SHARED.md, with bounded partitions and actual image inspection. Headless checks are allowed; do not open Matt's panels or leave task-owned tabs open.

## Constraints and rational changes
Do not build a backend, real authentication, storage service, live chat or scheduling. Do not rewrite Rita's exercises, regenerate 56 satisfactory visuals, or split app.js between competing owners.

If a verification criterion is objectively wrong, duplicated or tests the wrong outcome, record its original text, replacement, rationale and evidence in your report. Preserve the user's requested outcome and explicit constraints. An unfinished hard requirement cannot be converted into an optional one to claim DONE.

## Completion and blocked state
Write `/Users/agency/Documents/Agty/KomplexLogopédia/outputs/rita-ui-parallel-20261002/handoffs/04-course/manifest.json` using the small shared schema and your concise report at `/Users/agency/Documents/Agty/KomplexLogopédia/outputs/rita-ui-parallel-20261002/reports/04-course.md`, based on `/Users/agency/.agents/skills/launch-agent/report-template.md`. A manifest marked ready means this owned UI works and its local checks pass. The publisher performs the final deployed acceptance.

If no viable in-scope recovery remains, finish independent UI work and record the exact affected dependency, actual evidence and resume point. Do not manufacture blockers from optional polish. Return once in your final channel: DONE or STALLED, report path and the usable entry route. Native children use their normal completion channel back to parent `01a0c622-0f03-7480-9f24-003a441af4b3`. Separate visible tasks report through their own final response; do not send a second message to another task. No per-child wrap-up/audio.