LL2 · Matt’s communication template

Komplex · coherent journey rebuild

● IN PRODUCTION

Execute the authorized frontend review package autonomously and maintain evidence in this contract.

Live owner: GPT-6.1 Sol Ultra executor, 01a0f993-6786-74f3-9b94-3cbdb42e890f Run: KL-JOURNEY-20261001 Updated: 2026-10-01T23:28:18.537806+00:00

Authority: Matt’s 2 October instruction authorizes execution after reconciling all four browser comments. The frontend-only boundary is fixed. These are production requirements; a green check needs actual evidence. The parent goes idle after dispatch. Two-diagram decision map.

Feedback for this review: The comment layer is installed, but remote saving is temporarily unavailable because the shared Cloudflare service has reached its plan limit. Send plan changes in the Codex chat. Do not rely on unsaved comments.

Goal and finish line

End goal: Matt can follow the actual journey in one canvas, compare alternatives, comment on the outer layer, choose the recommended set and send Rita a precise review email. A future execution loop can activate the acquisition system and customer infrastructure.

Done means: Target: the fully linked, mobile-tested frontend canvas and complete evidence. Local production and interaction receipts are now attached. Counts and screenshots remain distinct from independent acceptance.

0 / 44criteria verified
9deliverables
3open decisions

Required package: 20 square ads + four landing variants + 14 participant days + 14 Rita review screens. The images are 20 ads + seven landing visuals + 56 daily alternatives = 83 actual generated visual assets.

Rules that stay fixed

  • Matt authorized a pinned GPT-6.1 Sol Ultra /goal executor. Build the frontend review package only.
  • Keep daily Rita feedback and original exercise wording. Customer interfaces and QA are mobile only.
  • Bare review URLs, preserved comment identity, no automatic opening. Shared external API limit: USD15.
Proof standard: Acceptance needs actual assets, coherent reader-to-next-step copy, source comparisons, working demo states and fresh deployed mobile screenshots. Each criterion records its current evidence and remaining verification.

Deliverable flow

Inputs → outputs · maximum eight nodes

The intended customer journey, with the current human-review gate kept explicit.

Source inputs

Rita, Figma and live identity

Why: Use six original modules, the full persona workbook, commented Figma variants and current Komplex design.

Thinking before design

One meta-persona

Why: Describe common situations and objections, then explain the next decision at each step.

Attention

20 new square ads

Why: Luna criticizes each caption before GPT imagery. The matching copy explains the offer and click.

Offer

Four complete landing variants

Why: Two variants per offer, all supplied layouts, seven new visuals and clear CTA alternatives.

Purchase and access

Stripe and email states

Why: Show the demo handoff, confirmation, access, nurture and daily delivery without real transactions.

Practice and feedback

14 participant + 14 Rita screens

Why: Exact exercise text, daily forms/video, 56 visual alternatives, previous-day review and daily feedback.

Assembled outcome

Canvas, proof and client email

Why: Every asset is linked in journey order, with expert screens in the row below and comments on the outside.

Current authority

GPT-6.1 Sol Ultra execution authorized

Matt authorized autonomous frontend delivery. The executor maintains the proof and prepares one morning handoff.

What changed in this contract

Latest instruction governs
Matt’s launch · 2 OctoberFrontend execution authorized

All four browser comments are reconciled: precise participant acceptance, frontend-only Rita and checkout prototypes, and the actual CC Klíma sales-call canvas. Launch one pinned GPT-6.1 Sol Ultra thread with /goal. Rational criterion replacements require a before/after log and equivalent or stronger proof. Parent goes idle.

Source and scope clarificationsSource snapshot

Figma recheck: 178 comments with unchanged IDs, messages and resolution states. The new kit has 28 labeled layouts, S1–S28. Rita’s source has six modules covering the full 14-day range. Daily reflection forms need saved readback, and payment success/access is a simulation rather than proof of real fulfillment.

Sources and plan decisionsSmall files first

These are inputs and proposed boundaries. They are not production evidence.

Latest explicit requirements: Canvas, 20 square ads, 4 landing variants, 7 landing visuals, 14 mobile participant days, 14 aligned Rita screens, 56 daily image alternatives and complete journey copy.

Rita source wording and Figma insight: Exercise wording remains authoritative. Earlier objections about workload or less feedback are superseded by Matt’s daily-feedback pilot. Fresh Figma REST read on 1 October returned 178 comments, unchanged in IDs, messages and resolution state from the preserved September snapshot. Original source contains six modules, not 14 separate lessons.

Current website identity: Use the live Komplex website and relevant Figma designs as visual evidence, plus all supplied new-template layouts.

Commercial details and dates: Carry the prior 24,900 Ft and 12-person pilot as a proposal. Start/close dates and real final-consultation booking URL remain unconfirmed.

Meeting, comments and later overrides: Re-read the latest located Rita meeting transcript before production, using find-transcript to resolve any newer authoritative record. The preserved full 18 August transcript is in .tmp/client-review-20260922/meeting-20260818-part1.txt. Keep client agreements, source text and later Matt proposals distinct. Latest daily-feedback and review-gate instructions take precedence.

Latest review comments and launch authority: Comment 1 requires much more precise participant acceptance. Comments 2 and 3 explicitly limit app, Rita and checkout to frontend prototypes. Comment 4 requires the previously supplied sales-call HTML canvas, https://review.clientsflow.hu/ccklima-canvas-2026-10-01-v1/. Matt now explicitly requests a new pinned GPT-6.1 Sol Ultra Codex thread with /goal, autonomous completion, logged rational criterion changes, then parent idle.

Launch is now explicitly authorized. One pinned GPT-6.1 Sol Ultra thread owns autonomous execution and verification. Parent goes idle after launch and startup acknowledgement, without ongoing supervision.

Source: Latest Matt instruction on 2 October: create a new Codex thread with GPT-6.1 Sol Ultra, give it /goal, then go idle.

Optimize customer interfaces for mobile only. Do not QA larger screen sizes.

Source: Latest user: Also create these interfaces optimized for mobile only. Don’t even QA bigger screen sizes.

Daily guides, daily participant videos and daily Rita feedback are the primary pilot. Ignore older low-workload objections for this proposal.

Source: Latest user and daily-feedback override dated 26 September.

Publish bare review URLs and do not automatically open files, panels or browser tabs.

Source: Latest AGENTS deliverable and HTML instructions.

Keep aggregate external API spending at or below USD15 for this task. Preserve receipts and a single cost ledger across workers.

Source: Latest AGENTS authority exception.

Write all customer-facing ads, landings, emails and application copy in natural Hungarian. Preserve Rita’s original Hungarian exercise wording and accents. Keep internal explanations separate. Avoid em dashes and semicolons in client-facing copy.

Source: Current Hungarian website, original Rita materials, PROJECT_STATE.md language invariant, and Matt’s client-copy instructions.

Build the complete frontend review prototypes only. Use fixtures and browser-local demonstration state. Do not implement backend APIs, real authentication, hosted video storage, Stripe payment creation/webhooks or email sending. Existing comment-host services may be reused. A prototype must still have coherent, meaningful content and connected interactions.

Source: Matt browser comments 2 and 3: we are only creating the frontend now; only need the frontend prototype; do not worry about the backend.

The executor may replace a verification criterion only for a concrete rational reason that better proves the original requested outcome. Before/after text, criterion ID, rationale, source/constraint provenance, alternative proof and acceptance impact must be logged in criterion-changes.md and reflected in the board. Do not weaken outcomes to hide failures, waive missing assets, substitute unrequested generation models, change hard quantities or claim an old check passed after replacing it. A parent-created interpretation may be revised when the original requirement remains satisfied.

Source: Latest Matt instruction: if it has a good reason it can replace some verification criteria, but must keep a log, only if really rational.

Keep criterion-changes.md beside the board and link it in the completed review canvas/handoff. Log before/after rationale and alternative evidence before applying a replacement. No superficial test can substitute for real source, content and UX proof.

Source: Matt’s latest explicit criterion-replacement permission, with rationality and change-log conditions.

GPT image generator for requested imagery, Gemini 3.1 Pro for the four landing generation calls, actual Luna critique and QA.

Canonical comment-html-review-host publisher, preserving shared-host routes and comment identity.

Use the user-requested Impeccable workflow for the landing and mobile UI redesign, preserving the established Komplex identity. Source and template mapping precede generation.

GPT-6.1 Sol Ultra, expressly replacing the earlier Astra High choice. Fresh /goal assignment, exclusive execution ownership, concise report. No parent progress polling.

Work and proof

One card per deliverable

Existing review comments and asset critique

Put comments on the September master and critique every existing asset.

AUTHORIZED0 / 2

Current problem: Review is difficult to scan and embedded comment layers fragment the discussion.

Proposed change: 267 asset-specific critiques are saved in the staged September master. The outer-only adapter preserves all child runtimes, identities and source bytes. Publication and provider persistence are checked separately.

Visible at: Existing September master and its asset cards

Checks and constraints

☐ Only the outer master shows the comment UI for embedded assets. Existing comments and inline edits are preserved.

Verification: Inspect actual deployed master, its embedded documents and comment readback. Test persistence in a disposable fixture.

Evidence and screenshot brief · 1 capture group(s)
local_evidence_ready. 267 granular critiques, 29 embeds and all137 original child files preserved. Local runtime/adapter fixtures pass. Actual hosted layer and provider readback pending.

Completed worker evidence and its explicit limits

Required screenshot evidence

Capture the state below at readable scale. Luna first gives a blind description, then checks the visible indicators. Judge all required coverage, not one attractive example.

1. One outer review layer, selected asset0 captured · 0 review record(s)
SCREENSHOT PLACEHOLDER · NOT CAPTUREDReplace with genuine postdeployment captures of the screen and state below.This is a capture brief, not proof or a passed check.
Capture
Existing September master, with a visible ad or landing card and the OUTER comment selection marker/composer.
Set up / action
Enable review mode on the master; select a specific asset card. Expand its embedded screen while keeping the outer asset identity visible.
Viewport
Internal review canvas/document: 1280×900 CSS px, DPR 1. Open customer screens separately at mobile widths. A canvas thumbnail does not replace a readable mobile screenshot.
Required coverage
Capture outer selection and expanded asset. Inspect every embedded document for duplicate toolbars; include one readable landing example.

What independent Luna must be able to see/read:

  • The selected asset has a stable identity and a precise outer selection/marker.
  • Only the master exposes the review toolbar/composer. The embedded customer interface remains unobstructed.
  • The marker refers to the intended asset or section, not an unrelated iframe boundary.
Visible reasons to reject: A second toolbar/comment composer appears inside the iframe. The selected target is ambiguous, or a review overlay hides the customer copy.

Evidence fields to fill: screenshot URL(s) · asset/day ID · state · viewport · capture time · deployed screen URL · actual Luna observations · reviewer report.

Also required. A screenshot cannot prove this:
  • Preserve the old document ID and saved edit/comment records. Disposable-fixture persistence and readback receipts are required. A toolbar screenshot does not prove remote saving. If the shared comment service is unavailable, disclose it and preserve records rather than marking persistence passed.

☐ Every existing review asset has a saved critique identifying overview difficulty, reader reasoning, a concrete weakness and an actionable communication change. Landings receive section-level critique. Count individual creatives and email variants as assets, not only the 29 embedded documents.

Verification: Compare the complete existing asset inventory and text with saved outer-layer comments and the critique index.

Evidence and screenshot brief · 1 capture group(s)
local_evidence_ready. 267 granular critiques, 29 embeds and all137 original child files preserved. Local runtime/adapter fixtures pass. Actual hosted layer and provider readback pending.

Completed worker evidence and its explicit limits

Required screenshot evidence

Capture the state below at readable scale. Luna first gives a blind description, then checks the visible indicators. Judge all required coverage, not one attractive example.

1. Asset and concrete audience critique together0 captured · 0 review record(s)
SCREENSHOT PLACEHOLDER · NOT CAPTUREDReplace with genuine postdeployment captures of the screen and state below.This is a capture brief, not proof or a passed check.
Capture
Old master asset cards and section-level landing critique.
Set up / action
Expand the existing asset and its saved critique at readable scale.
Viewport
Internal review canvas/document: 1280×900 CSS px, DPR 1. Open customer screens separately at mobile widths. A canvas thumbnail does not replace a readable mobile screenshot.
Required coverage
For every inventory asset attach asset-ID-linked screenshots. Landings require each section-level critique and the section it refers to.

What independent Luna must be able to see/read:

  • The actual creative, email or landing passage being criticized is readable.
  • Critique identifies why overview is difficult, the likely reader interpretation, a specific weakness and a concrete replacement/change.
  • The suggested change concerns this asset and its next action; it is not a generic CRO checklist.
Visible reasons to reject: Only a summary says assets were reviewed. Critique could apply unchanged to any asset, or the relevant copy is too small to read.

Evidence fields to fill: screenshot URL(s) · asset/day ID · state · viewport · capture time · deployed screen URL · actual Luna observations · reviewer report.

Also required. A screenshot cannot prove this:
  • Compare the critique index with the full inventory, including each individual ad and email variant. Save exact comment/critique text and target IDs. Counts alone do not prove an audience-aware critique.
Shared rules and source boundaries above apply. The check state stays pending until evidence exists.

Audience and persuasion logic

Define one broad meta-persona and explain the reader’s decision at every funnel step.

AUTHORIZED0 / 2

Current problem: Generic hooks and disconnected feature lists do not explain why the next action matters to this visitor.

Proposed change: Use the existing persona workbook and real source insights to define common problems, self-description, everyday situations and feelings. For each asset record attention, interpretation order, internal narrative, objections and how the next step earns its place. Apply the Value Equation qualitatively where useful, without invented scores. Keep analysis outside the customer screen.

Visible at: Canvas introduction and concise analysis cards beside each stage

Checks and constraints

☐ One broad adult meta-persona uses the workbook’s frame and distinguishes supported observations from hypotheses. It covers problem, self-framing, everyday situations, feelings and desired progress without overcomplicated targeting segments. Include a clear proposed Meta Advantage+ broad-targeting setup, with geography, placements and test assumptions to verify in the later campaign loop. Do not imply any account setting was activated.

Verification: Read the complete dynamic workbook including app.js and the final commonalities. Compare claims to source evidence and the current broad-audience pilot.

Evidence and screenshot brief · 1 capture group(s)
pending. Evidence pending. Do not mark passed until the actual required output and verification exist.
Required screenshot evidence

Capture the state below at readable scale. Luna first gives a blind description, then checks the visible indicators. Judge all required coverage, not one attractive example.

1. Meta-persona and proposed targeting0 captured · 0 review record(s)
SCREENSHOT PLACEHOLDER · NOT CAPTUREDReplace with genuine postdeployment captures of the screen and state below.This is a capture brief, not proof or a passed check.
Capture
Canvas introduction: audience commonalities plus proposed Meta targeting card.
Set up / action
Show the complete persona and targeting assumptions; expand any hidden text.
Viewport
Internal review canvas/document: 1280×900 CSS px, DPR 1. Open customer screens separately at mobile widths. A canvas thumbnail does not replace a readable mobile screenshot.
Required coverage
One complete persona capture set, including common problem, self-framing, situations, emotions, desired progress and targeting proposal.

What independent Luna must be able to see/read:

  • Concrete everyday situations explain who the broad adult audience is and what they want to improve.
  • Problem, self-narrative, feelings and desired progress are distinct and understandable, with hypotheses labelled.
  • Targeting card clearly states proposed geography, broad/Advantage+ approach, placements and test assumptions; no claim of campaign activation.
Visible reasons to reject: Vague segments replace usable commonalities. Unsupported psychological facts, invented clinical labels or activated-account claims appear.

Evidence fields to fill: screenshot URL(s) · asset/day ID · state · viewport · capture time · deployed screen URL · actual Luna observations · reviewer report.

Also required. A screenshot cannot prove this:
  • Record workbook source passages, shared observations and hypotheses separately. Read the dynamic workbook app.js. Proposed targeting must not be described as activated Meta settings.

☐ Before visual generation, each funnel step has a saved Markdown copy draft and simple wireframe, plus a concrete criticism that is reflected in the final asset. Each major step then has concise audience-reasoning cards covering before/attention/interpretation/order/objection/internal narrative/desired next thought and a communication improvement. Desired benefit, believable delivery and friction are considered when relevant.

Verification: Read the saved drafts, wireframes and criticisms for each step. Compare the final asset and each transition to the reasoning cards. Read hook, caption, body copy and CTA in order. Treat resulting judgments as hypotheses to test.

Evidence and screenshot brief · 1 capture group(s)
pending. Evidence pending. Do not mark passed until the actual required output and verification exist.
Required screenshot evidence

Capture the state below at readable scale. Luna first gives a blind description, then checks the visible indicators. Judge all required coverage, not one attractive example.

1. Reader reasoning beside the final step0 captured · 0 review record(s)
SCREENSHOT PLACEHOLDER · NOT CAPTUREDReplace with genuine postdeployment captures of the screen and state below.This is a capture brief, not proof or a passed check.
Capture
Canvas analysis card next to the actual ad, landing, email, checkout, lesson, feedback or consultation screen.
Set up / action
Open final asset and its reasoning card. Show hook/caption/body/CTA in actual reading order.
Viewport
Internal review canvas/document: 1280×900 CSS px, DPR 1. Open customer screens separately at mobile widths. A canvas thumbnail does not replace a readable mobile screenshot.
Required coverage
Every major step on both entry routes and the daily/consultation journey has a linked reasoning capture, not just the persona section.

What independent Luna must be able to see/read:

  • Card distinguishes before-view attitude, attention hook, interpretation order, internal narrative and objection.
  • It states the desired next thought and explains why the next action offers relevant value with credible delivery and manageable effort.
  • A concrete criticism and resulting copy/UX improvement match what is visible in the final asset.
Visible reasons to reject: Cards merely list USPs or score imagined customer beliefs. Reasoning is mixed into the customer-facing UI, or the final asset ignores the stated objection.

Evidence fields to fill: screenshot URL(s) · asset/day ID · state · viewport · capture time · deployed screen URL · actual Luna observations · reviewer report.

Also required. A screenshot cannot prove this:
  • Keep pre-generation Markdown drafts, wireframes, criticisms and revisions for every funnel step. Record chronology and final-source mapping. A screenshot cannot prove drafts preceded visual generation or predict the reader’s actual thoughts.
Shared rules and source boundaries above apply. The check state stays pending until evidence exists.

One left-to-right customer journey canvas

Show every asset in customer order, with Rita feedback directly below each daily screen.

AUTHORIZED0 / 4

Current problem: The current long document buries the sequence and makes variants difficult to compare.

Proposed change: Adapt the CC Klíma pan/zoom canvas into a coherent two-row journey. Provide a recommended path and selectable alternatives, direct asset links, a complete index and outer commenting.

Visible at: New public canvas

Checks and constraints

☐ Ads, both entry routes, free upload/feedback, consent and email states, paid landing, Stripe states, confirmation/access, all 14 days, Rita review, final video and consultation appear in logical order.

Verification: Follow both normal entry paths through the actual canvas and each linked screen. Check one-to-one alignment of 14 Rita screens below the matching participant days.

Evidence and screenshot brief · 1 capture group(s)
pending. Evidence pending. Do not mark passed until the actual required output and verification exist.
Required screenshot evidence

Capture the state below at readable scale. Luna first gives a blind description, then checks the visible indicators. Judge all required coverage, not one attractive example.

1. Both entry routes through consultation0 captured · 0 review record(s)
SCREENSHOT PLACEHOLDER · NOT CAPTUREDReplace with genuine postdeployment captures of the screen and state below.This is a capture brief, not proof or a passed check.
Capture
Fitted canvas overview plus readable acquisition and daily-row crops.
Set up / action
Use Fit, then focus the acquisition branch and the 14-day participant/Rita rows.
Viewport
Internal review canvas/document: 1280×900 CSS px, DPR 1. Open customer screens separately at mobile widths. A canvas thumbnail does not replace a readable mobile screenshot.
Required coverage
Capture full overview, both entry branches and row alignment covering every day-01–day-14.

What independent Luna must be able to see/read:

  • Free-video and direct-challenge routes join the correct paid/access journey.
  • Upload/feedback, consent, emails, checkout, confirmation/access, all daily screens, final video and consultation appear in logical order.
  • Exactly 14 expert cards sit below matching participant day IDs, with recognizable labels.
Visible reasons to reject: A route jumps directly from ad to a lesson without explaining access. Days are missing, out of order, or expert and participant IDs do not align.

Evidence fields to fill: screenshot URL(s) · asset/day ID · state · viewport · capture time · deployed screen URL · actual Luna observations · reviewer report.

Also required. A screenshot cannot prove this:
  • Click through both acquisition paths and compare actual destinations/day IDs. Retain transition receipts and the complete manifest. Connector arrows alone do not prove the journey works.

☐ The canvas has pan, zoom, fit, jump navigation, expand/view controls and a clear recommended route. All alternatives and decisions can be reviewed without losing the journey. Comments are on the outer layer.

Verification: Use the deployed controls, inspect focus/keyboard behavior and verify every asset is reachable from the complete index.

Evidence and screenshot brief · 1 capture group(s)
pending. Evidence pending. Do not mark passed until the actual required output and verification exist.
Required screenshot evidence

Capture the state below at readable scale. Luna first gives a blind description, then checks the visible indicators. Judge all required coverage, not one attractive example.

1. Readable canvas and full-screen inspection0 captured · 0 review record(s)
SCREENSHOT PLACEHOLDER · NOT CAPTUREDReplace with genuine postdeployment captures of the screen and state below.This is a capture brief, not proof or a passed check.
Capture
Canvas at normal reading zoom and an expanded asset view.
Set up / action
Jump to an asset, open its readable full view, then return to the same canvas position.
Viewport
Internal review canvas/document: 1280×900 CSS px, DPR 1. Open customer screens separately at mobile widths. A canvas thumbnail does not replace a readable mobile screenshot.
Required coverage
Capture controls, focused stage, expanded view and restored position.

What independent Luna must be able to see/read:

  • Pan/zoom/Fit/jump/open controls have understandable labels and the recommended route is obvious.
  • Full copy is readable in an expanded view without tiny thumbnails being treated as the review surface.
  • Outer comments and alternative-selection controls do not hide navigation.
Visible reasons to reject: The canvas is only a very wide static image. Reviewer must lose position or navigate away blindly to read an asset.

Evidence fields to fill: screenshot URL(s) · asset/day ID · state · viewport · capture time · deployed screen URL · actual Luna observations · reviewer report.

Also required. A screenshot cannot prove this:
  • Exercise pan, zoom, fit, jump, expand, return, keyboard/focus and every index link. Record the same viewport position before/after opening a screen.

☐ Adapt the actual user-supplied sales-call canvas HTML at https://review.clientsflow.hu/ccklima-canvas-2026-10-01-v1/ or its saved source. Retain its usable spatial/pan/zoom/jump/open-screen behaviour. Do not replace it with another long vertical report or only a Mermaid chart.

Verification: Save a concise source-to-new-canvas component mapping. In the deployed review use the real controls to traverse acquisition to consultation, inspect a full screen and return to the same position. Verify the displayed source template and the actual output, not a visual claim alone.

Evidence and screenshot brief · 1 capture group(s)
pending. Evidence pending. Do not mark passed until the actual required output and verification exist.
Required screenshot evidence

Capture the state below at readable scale. Luna first gives a blind description, then checks the visible indicators. Judge all required coverage, not one attractive example.

1. Supplied canvas behavior retained0 captured · 0 review record(s)
SCREENSHOT PLACEHOLDER · NOT CAPTUREDReplace with genuine postdeployment captures of the screen and state below.This is a capture brief, not proof or a passed check.
Capture
Source CC Klíma canvas and the new customer journey canvas.
Set up / action
Capture matching control areas and the new focused journey view; use the actual source HTML, not a recreated drawing.
Viewport
Internal review canvas/document: 1280×900 CSS px, DPR 1. Open customer screens separately at mobile widths. A canvas thumbnail does not replace a readable mobile screenshot.
Required coverage
One source/output comparison plus open-and-return state of the new canvas.

What independent Luna must be able to see/read:

  • New page retains useful spatial navigation, pan/zoom/jump and screen-opening behavior.
  • The primary structure is a left-to-right journey with expert feedback below participant days.
  • Branding and content are Komplex, while the reusable canvas navigation remains recognizable.
Visible reasons to reject: The output is another vertical report or only a Mermaid diagram. Decorative canvas controls are visible but not usable.

Evidence fields to fill: screenshot URL(s) · asset/day ID · state · viewport · capture time · deployed screen URL · actual Luna observations · reviewer report.

Also required. A screenshot cannot prove this:
  • Save a source-to-output component mapping and actual control behavior receipt from the supplied CC Klíma canvas. Visual resemblance alone does not prove template use.

☐ A new reviewer can find the recommended complete path, all ad/landing/email alternatives, each day’s participant and Rita views, critiques, important AI decisions and the client email from one assembled canvas/index. Full text is available without unreadably shrinking the interfaces.

Verification: Start from the canvas introduction and follow its stated review order. Independently locate one asset from every deliverable and compare A/B alternatives side by side. Verify every manifest item is reachable and the second-row Rita days align to the same 14 participant day IDs.

Evidence and screenshot brief · 1 capture group(s)
pending. Evidence pending. Do not mark passed until the actual required output and verification exist.
Required screenshot evidence

Capture the state below at readable scale. Luna first gives a blind description, then checks the visible indicators. Judge all required coverage, not one attractive example.

1. Recommended path and A/B comparisons0 captured · 0 review record(s)
SCREENSHOT PLACEHOLDER · NOT CAPTUREDReplace with genuine postdeployment captures of the screen and state below.This is a capture brief, not proof or a passed check.
Capture
Canvas introduction/review index and a readable side-by-side variant panel.
Set up / action
Follow the stated review order; open an ad, both offer alternatives, one email and paired participant/Rita day.
Viewport
Internal review canvas/document: 1280×900 CSS px, DPR 1. Open customer screens separately at mobile widths. A canvas thumbnail does not replace a readable mobile screenshot.
Required coverage
Capture introduction, complete index groups, comparison state and matched day rows.

What independent Luna must be able to see/read:

  • Recommended selections are labelled as recommendations, with alternatives easy to locate.
  • Every deliverable group, important AI decision, critique and unsent client email has a visible route/link.
  • Side-by-side alternatives remain readable and the user can return to the journey.
Visible reasons to reject: Variants are buried in unlabelled iframe tabs. Decisions and full copy can only be found by searching the HTML source.

Evidence fields to fill: screenshot URL(s) · asset/day ID · state · viewport · capture time · deployed screen URL · actual Luna observations · reviewer report.

Also required. A screenshot cannot prove this:
  • Perform an independent find-one-item-per-deliverable review and compare every manifest item to a reachable index entry. Capture the reviewer’s confusion and corrections, not an unsupported claim of fast review.
Shared rules and source boundaries above apply. The check state stays pending until evidence exists.

20 new square ads with natural Hungarian copy

Replace weak captions with 20 coherent, image-led Facebook concepts.

AUTHORIZED0 / 3

Current problem: Existing copy spends attention on service details without sufficiently locating the reader in a relevant situation.

Proposed change: Proposed split: 10 free-video and 10 direct-challenge ad concepts, each with real 1:1 imagery and matching full Facebook copy. Independent Luna critiques each proposed caption and its narrative before image generation.

Visible at: Canvas ad area and full Facebook previews

Checks and constraints

☐ 20 distinct new GPT image outputs are square and contain the intended natural copy. No ad mentions Rita by name.

Verification: Verify model receipts, dimensions, all final images and the exact rendered caption against final approved copy.

Evidence and screenshot brief · 1 capture group(s)
pending. Evidence pending. Do not mark passed until the actual required output and verification exist.
Required screenshot evidence

Capture the state below at readable scale. Luna first gives a blind description, then checks the visible indicators. Judge all required coverage, not one attractive example.

1. All 20 finished square Facebook creatives0 captured · 0 review record(s)
SCREENSHOT PLACEHOLDER · NOT CAPTUREDReplace with genuine postdeployment captures of the screen and state below.This is a capture brief, not proof or a passed check.
Capture
Full Facebook preview for each ad-01–ad-20, including actual image and matching copy.
Set up / action
Open each ad at readable mobile scale; expand long primary text if needed.
Viewport
Customer screen at 390×844 CSS px, DPR 1 for readable content. Add 360×800 and 430×932 for the stated mobile QA coverage. Never capture a desktop customer UI.
Required coverage
20 individual image/copy captures. Inspect all final image originals for square dimensions and caption spelling.

What independent Luna must be able to see/read:

  • Each image is a genuine rendered 1:1 creative, not a placeholder or stretched crop.
  • Caption is natural Hungarian, legible and relevant to its image and offer.
  • Rita’s name is absent from every ad. The primary text makes clear what the click will open.
Visible reasons to reject: Unreadable or misspelled image text, unexpected name, duplicated concept, placeholder media or offer mismatch appears.

Evidence fields to fill: screenshot URL(s) · asset/day ID · state · viewport · capture time · deployed screen URL · actual Luna observations · reviewer report.

Also required. A screenshot cannot prove this:
  • Verify 20 distinct new GPT-generation receipts, image dimensions and file hashes. Compare exact image text against final copy. A screenshot does not establish model provenance or file uniqueness.

☐ Each creative has pre-generation Luna criticism, final revised caption, matching primary text, headline, CTA and destination. The hook, image and offer form an understandable sequence.

Verification: Read all 20 independent critique records and final ad packs. Inspect actual mobile previews at customer scale.

Evidence and screenshot brief · 1 capture group(s)
pending. Evidence pending. Do not mark passed until the actual required output and verification exist.
Required screenshot evidence

Capture the state below at readable scale. Luna first gives a blind description, then checks the visible indicators. Judge all required coverage, not one attractive example.

1. Coherent image → copy → destination0 captured · 0 review record(s)
SCREENSHOT PLACEHOLDER · NOT CAPTUREDReplace with genuine postdeployment captures of the screen and state below.This is a capture brief, not proof or a passed check.
Capture
Each final Facebook preview plus its destination landing’s first screen.
Set up / action
Read the image/caption first, then primary text/headline/CTA; open its declared destination.
Viewport
Customer screen at 390×844 CSS px, DPR 1 for readable content. Add 360×800 and 430×932 for the stated mobile QA coverage. Never capture a desktop customer UI.
Required coverage
All 20 ad packs and both offer destinations; attach critique record IDs outside the customer preview.

What independent Luna must be able to see/read:

  • Hook locates the reader in a recognizable situation or clear curiosity.
  • Image caption and primary text form one understandable thought, then explain the offer and next step.
  • The destination’s first screen answers the expectation created by the ad.
Visible reasons to reject: The body is an unexplained USP list. The headline changes subject, or a free-video ad opens a paid offer without the promised step.

Evidence fields to fill: screenshot URL(s) · asset/day ID · state · viewport · capture time · deployed screen URL · actual Luna observations · reviewer report.

Also required. A screenshot cannot prove this:
  • Preserve pre-generation Luna critiques and before/after text for all 20 concepts. Receipts must establish critique before the generated image, not a retrospective rubber stamp.

☐ Angles differ materially. A concise test guide identifies what to compare, which entry offer it supports and the downstream outcome to watch.

Verification: Read the test guide and check each ad destination/message match.

Evidence and screenshot brief · 1 capture group(s)
pending. Evidence pending. Do not mark passed until the actual required output and verification exist.
Required screenshot evidence

Capture the state below at readable scale. Luna first gives a blind description, then checks the visible indicators. Judge all required coverage, not one attractive example.

1. Materially different ad angles0 captured · 0 review record(s)
SCREENSHOT PLACEHOLDER · NOT CAPTUREDReplace with genuine postdeployment captures of the screen and state below.This is a capture brief, not proof or a passed check.
Capture
Canvas ad test guide with the corresponding creative/copy pairs.
Set up / action
Show representative variants for each angle, alongside what the comparison is intended to teach.
Viewport
Internal review canvas/document: 1280×900 CSS px, DPR 1. Open customer screens separately at mobile widths. A canvas thumbnail does not replace a readable mobile screenshot.
Required coverage
Every ad ID maps to a test angle and destination. Capture the guide and readable comparisons for the distinct angles.

What independent Luna must be able to see/read:

  • Variants differ in a clear situation, hook, benefit emphasis or visual concept, not only a swapped adjective.
  • Each proposed comparison identifies the audience/offer and a downstream action to observe.
  • Recommended starting set is distinct from measured results.
Visible reasons to reject: Twenty nearly identical ads are presented as twenty tests. The plan optimizes clicks without considering whether visitors submit or join.

Evidence fields to fill: screenshot URL(s) · asset/day ID · state · viewport · capture time · deployed screen URL · actual Luna observations · reviewer report.

Also required. A screenshot cannot prove this:
  • Keep the test matrix with concept IDs, entry offers, comparison hypotheses and downstream metrics. Do not label simulated or untested variants measured winners.
Shared rules and source boundaries above apply. The check state stays pending until evidence exists.

Four new landing variants and seven custom visuals

Build two complete variants per offer using all supplied template layouts.

AUTHORIZED0 / 5

Current problem: The existing landing variants are lengthy and their persuasion, hierarchy and message match need improvement.

Proposed change: Use the live Komplex visual identity, current Rita Figma insight and the new_template-3.html layout vocabulary. Each offer gets distinct A/B routes and three main CTA-section alternatives. Use Gemini 3.1 Pro for the four landing generations and GPT image models for seven custom visuals.

Visible at: Four standalone landing URLs and canvas comparisons

Checks and constraints

☐ All 28 labeled section layouts S1–S28 from new_template-3.html are meaningfully adapted across the four-page family, with a documented source-to-page mapping and a distinct reader decision for each layout. Each of the four pages is complete, coherent and materially different from its matching A/B alternative. Preserve recognizable layout structures, replace irrelevant functions with truthful offer-related content, and avoid repeated filler. No placeholders, fake proof or unsupported guarantees remain.

Verification: Create a template-to-page layout mapping. Compare deployed mobile sections to the supplied layout structures and current website/Figma visual references.

Evidence and screenshot brief · 1 capture group(s)
local_evidence_ready. Four actual Gemini3.1Pro complete generations and50 meaningful sections across all28 supplied layouts. Source mapping, seven generated visual references, CTA alternatives and technical mobile interaction receipts exist. Independent deployed mobile screenshots pending.

Completed worker evidence and its explicit limits

Required screenshot evidence

Capture the state below at readable scale. Luna first gives a blind description, then checks the visible indicators. Judge all required coverage, not one attractive example.

1. Actual layouts across four mobile landings0 captured · 0 review record(s)
SCREENSHOT PLACEHOLDER · NOT CAPTUREDReplace with genuine postdeployment captures of the screen and state below.This is a capture brief, not proof or a passed check.
Capture
free-video-A/B and challenge-A/B: ordered, readable section screenshot sets.
Set up / action
Use real deployed pages. Scroll so complete headings, copy, media and CTAs of each adapted section are visible.
Viewport
Customer screen at 390×844 CSS px, DPR 1 for readable content. Add 360×800 and 430×932 for the stated mobile QA coverage. Never capture a desktop customer UI.
Required coverage
All four pages, every adapted section, at 390; include 360/430 coverage in landing QA. Mapping must account for all requested S1–S28 layouts.

What independent Luna must be able to see/read:

  • Each section preserves an identifiable supplied layout while serving a distinct offer-related purpose.
  • Typography, palette, buttons and image treatment align with current Komplex references.
  • Every section uses finished relevant content, with honest proof and no fake testimonial or arbitrary filler.
Visible reasons to reject: Layout quota is met by repeating the same claim. Irrelevant pricing/features, fake proof, missing media or unrelated template content remain.

Evidence fields to fill: screenshot URL(s) · asset/day ID · state · viewport · capture time · deployed screen URL · actual Luna observations · reviewer report.

Also required. A screenshot cannot prove this:
  • Keep S1–S28 source-to-page mapping for all four pages, design-token references and a claim/source audit. If rationally amended, log original/revised layout coverage and justification. A screenshot cannot prove all layouts or truthful claims by itself.

☐ The first screen identifies the offer, who it is for, practical benefit and next action. Sections earn the next step and address the relevant objections without mixing the two offers. A/B variants and critical CTA alternatives have material testable differences. Include three full main CTA-section alternatives per offer and show them side by side for review.

Verification: Read every section as a new visitor, check all claims/source trace and click through every principal CTA to its proper demo state.

Evidence and screenshot brief · 2 capture group(s)
local_evidence_ready. Four actual Gemini3.1Pro complete generations and50 meaningful sections across all28 supplied layouts. Source mapping, seven generated visual references, CTA alternatives and technical mobile interaction receipts exist. Independent deployed mobile screenshots pending.

Completed worker evidence and its explicit limits

Required screenshot evidence

Capture the state below at readable scale. Luna first gives a blind description, then checks the visible indicators. Judge all required coverage, not one attractive example.

1. Four understandable first screens0 captured · 0 review record(s)
SCREENSHOT PLACEHOLDER · NOT CAPTUREDReplace with genuine postdeployment captures of the screen and state below.This is a capture brief, not proof or a passed check.
Capture
First mobile viewport of free-video-A, free-video-B, challenge-A and challenge-B.
Set up / action
Open as a first-time visitor, no scrolling. Hide reviewer overlays only, not customer content.
Viewport
Customer screen at 390×844 CSS px, DPR 1 for readable content. Add 360×800 and 430×932 for the stated mobile QA coverage. Never capture a desktop customer UI.
Required coverage
One first-screen capture per page at 390, plus 360/430 mobile QA.

What independent Luna must be able to see/read:

  • Offer, intended visitor, practical benefit and immediate action are understandable from the first screen.
  • Free-video route offers a clear first step; challenge route explains the two-week paid practice offer.
  • Headline, subhead and CTA work together, with no unexplained promise or technical filler.
Visible reasons to reject: The screen describes the provider before explaining why the visitor should care. Ambiguous CTA, mixed offers, obscure headline or unsupported outcome claim appears.

Evidence fields to fill: screenshot URL(s) · asset/day ID · state · viewport · capture time · deployed screen URL · actual Luna observations · reviewer report.

2. Three main CTA sections per offer0 captured · 0 review record(s)
SCREENSHOT PLACEHOLDER · NOT CAPTUREDReplace with genuine postdeployment captures of the screen and state below.This is a capture brief, not proof or a passed check.
Capture
Side-by-side comparison of three full CTA-section alternatives for each of two offers.
Set up / action
Show full section headline, supporting copy, inclusion/friction information and button, then the linked demo state.
Viewport
Customer screen at 390×844 CSS px, DPR 1 for readable content. Add 360×800 and 430×932 for the stated mobile QA coverage. Never capture a desktop customer UI.
Required coverage
Six full CTA alternatives, readable mobile renderings and corresponding destination first screen.

What independent Luna must be able to see/read:

  • Alternatives make materially different, coherent cases for the same next step.
  • Button label and surrounding copy explain what happens after the click.
  • Paid CTAs state the proposed offer/amount honestly; free CTAs do not imply payment.
Visible reasons to reject: Only button colors change. CTA contradicts nearby copy or hides the expected checkout/upload action.

Evidence fields to fill: screenshot URL(s) · asset/day ID · state · viewport · capture time · deployed screen URL · actual Luna observations · reviewer report.

Also required. A screenshot cannot prove this:
  • Read complete copy and sources, then click every principal CTA. Persuasion is a testable hypothesis, not proven conversion. Save material A/B differences and destination receipts.

☐ Seven custom GPT-generated illustrations/infographics are actually used on the landing pages.

Verification: Check provider receipts, image files, legibility, captions and deployed usage.

Evidence and screenshot brief · 1 capture group(s)
local_evidence_ready. Four actual Gemini3.1Pro complete generations and50 meaningful sections across all28 supplied layouts. Source mapping, seven generated visual references, CTA alternatives and technical mobile interaction receipts exist. Independent deployed mobile screenshots pending.

Completed worker evidence and its explicit limits

Required screenshot evidence

Capture the state below at readable scale. Luna first gives a blind description, then checks the visible indicators. Judge all required coverage, not one attractive example.

1. Seven landing visuals used in context0 captured · 0 review record(s)
SCREENSHOT PLACEHOLDER · NOT CAPTUREDReplace with genuine postdeployment captures of the screen and state below.This is a capture brief, not proof or a passed check.
Capture
Each custom illustration/infographic within its actual landing section.
Set up / action
Capture media at readable size with the accompanying explanatory heading/copy.
Viewport
Customer screen at 390×844 CSS px, DPR 1 for readable content. Add 360×800 and 430×932 for the stated mobile QA coverage. Never capture a desktop customer UI.
Required coverage
Seven unique visuals; include every placement and inspect each original at readable resolution.

What independent Luna must be able to see/read:

  • Illustration explains the adjacent offer/process instead of adding unrelated decoration.
  • Infographic labels are legible Hungarian, accurate and consistent with the written process.
  • Visual is uncropped where its meaning depends on labels; essential instructions remain in readable text.
Visible reasons to reject: Fake anatomy, garbled labels, irrelevant imagery or hidden instructions appear.

Evidence fields to fill: screenshot URL(s) · asset/day ID · state · viewport · capture time · deployed screen URL · actual Luna observations · reviewer report.

Also required. A screenshot cannot prove this:
  • Verify seven GPT-generation receipts, final file identities/dimensions and deployed placement mapping. Text or anatomy claims need source validation.

☐ The four pages visibly use the current Komplex typography, palette, buttons, shape/image style and Rita Figma insight. Each section has a purpose in the reader’s decision. Hero, three CTA alternatives per offer, programme/feedback details, practical concerns and proof use truthful coherent content.

Verification: Compare actual mobile sections to the live website, multiple relevant commented Figma variants and the supplied layout kit. Keep a design-token and section-purpose mapping. Read the complete page as a first-time visitor, then save independent Luna findings and corrected postdeployment screenshots.

Evidence and screenshot brief · 1 capture group(s)
local_evidence_ready. Four actual Gemini3.1Pro complete generations and50 meaningful sections across all28 supplied layouts. Source mapping, seven generated visual references, CTA alternatives and technical mobile interaction receipts exist. Independent deployed mobile screenshots pending.

Completed worker evidence and its explicit limits

Required screenshot evidence

Capture the state below at readable scale. Luna first gives a blind description, then checks the visible indicators. Judge all required coverage, not one attractive example.

1. Komplex identity, not a generic redesign0 captured · 0 review record(s)
SCREENSHOT PLACEHOLDER · NOT CAPTUREDReplace with genuine postdeployment captures of the screen and state below.This is a capture brief, not proof or a passed check.
Capture
Current website/commented Figma reference crops beside final hero, programme, feedback, proof and CTA sections.
Set up / action
Capture references without modifying them and corresponding deployed sections at readable scale.
Viewport
Internal review canvas/document: 1280×900 CSS px, DPR 1. Open customer screens separately at mobile widths. A canvas thumbnail does not replace a readable mobile screenshot.
Required coverage
Reference set plus these key section types across all four variants; six CTA alternatives included.

What independent Luna must be able to see/read:

  • Type hierarchy, palette, button shapes and imagery form the same recognizable visual system.
  • Rita’s meaningful copy insight is reflected without importing superseded low-feedback promises.
  • Sections clarify programme, daily feedback, practical concerns and next action with consistent tone.
Visible reasons to reject: A polished but unrelated visual identity replaces the website style. Older contradictory promises survive in one variant.

Evidence fields to fill: screenshot URL(s) · asset/day ID · state · viewport · capture time · deployed screen URL · actual Luna observations · reviewer report.

Also required. A screenshot cannot prove this:
  • Keep website/Figma design references, token mapping and section-purpose notes. Read relevant Rita comments and distinguish superseded workload feedback. Screenshot resemblance does not prove source consultation.

☐ Landing quality requires actual mobile section screenshots and a content/UX review, not merely Gemini provenance, section count or absence of console errors. There is no placeholder media, fake testimonial, irrelevant section or repeated claim padding the layout quota.

Verification: At 360/390/430 run the requested section QA, blind review, Luna fixes and fresh postdeployment re-QA. Record each significant defect and its resolution. If a repetitive template criterion is rationally amended, map all requested layouts to meaningful coverage and retain the original/changed criterion in the change log.

Evidence and screenshot brief · 2 capture group(s)
local_evidence_ready. Four actual Gemini3.1Pro complete generations and50 meaningful sections across all28 supplied layouts. Source mapping, seven generated visual references, CTA alternatives and technical mobile interaction receipts exist. Independent deployed mobile screenshots pending.

Completed worker evidence and its explicit limits

Required screenshot evidence

Capture the state below at readable scale. Luna first gives a blind description, then checks the visible indicators. Judge all required coverage, not one attractive example.

1. All landing sections at the required mobile widths0 captured · 0 review record(s)
SCREENSHOT PLACEHOLDER · NOT CAPTUREDReplace with genuine postdeployment captures of the screen and state below.This is a capture brief, not proof or a passed check.
Capture
Actual deployed sections of all four landing pages at 360/390/430.
Set up / action
Capture readable section runs including sticky CTA, long paragraphs, image labels, forms and validation.
Viewport
Customer screen at 390×844 CSS px, DPR 1 for readable content. Add 360×800 and 430×932 for the stated mobile QA coverage. Never capture a desktop customer UI.
Required coverage
All four variants at all three mobile widths. No section may be omitted; a very tall scaled-down screenshot is not readable evidence.

What independent Luna must be able to see/read:

  • Copy, headings and image labels are readable without clipping or a wall of repeated claims.
  • Controls do not overlap content or force horizontal scrolling.
  • Every section has a relevant job and every media asset is finished.
Visible reasons to reject: A desktop capture is substituted. Overflow, repeated filler, clipped CTA, illegible image text or missing section remains.

Evidence fields to fill: screenshot URL(s) · asset/day ID · state · viewport · capture time · deployed screen URL · actual Luna observations · reviewer report.

2. Material landing defects fixed and rechecked0 captured · 0 review record(s)
SCREENSHOT PLACEHOLDER · NOT CAPTUREDReplace with genuine postdeployment captures of the screen and state below.This is a capture brief, not proof or a passed check.
Capture
Same landing section/state before and after a correction, on actual deployed revision.
Set up / action
Use the same viewport, section and scroll context. Attach final URL/build identifier.
Viewport
Customer screen at 390×844 CSS px, DPR 1 for readable content. Add 360×800 and 430×932 for the stated mobile QA coverage. Never capture a desktop customer UI.
Required coverage
Every significant Luna/content defect gets a comparable before/after pair and fresh independent observation.

What independent Luna must be able to see/read:

  • The exact reported problem is visible in the before image and absent in the after image.
  • Correction retains coherent offer and design, with surrounding context still readable.
Visible reasons to reject: Only an unrelated screenshot or a green status badge is supplied. After image is stale, local-only or at a different scale that hides the defect.

Evidence fields to fill: screenshot URL(s) · asset/day ID · state · viewport · capture time · deployed screen URL · actual Luna observations · reviewer report.

Also required. A screenshot cannot prove this:
  • Attach independent Luna section findings, correction receipts and fresh postdeployment QA. No missing-console-error or Gemini receipt can substitute for visual/content review.
Shared rules and source boundaries above apply. The check state stays pending until evidence exists.

Mobile participant app and fourteen full daily lessons

Show login, practice, reflection, video submission and yesterday’s feedback in a usable mobile app.

AUTHORIZED0 / 12

Current problem: Current daily material is a reading presentation rather than a clear participant task flow.

Proposed change: Deliver a coherent mobile frontend: demo entry/login, course overview, 14 complete day views, practice, original reflection forms, video preview/submission and feedback readback. Preserve Rita’s original exercise wording and source ranges. Two infographics and two AI-image alternatives per day. No backend implementation.

Visible at: Mobile app, 14 participant day cards and image chooser

Checks and constraints

☐ All 14 days retain Rita’s original exercise wording and meaning. Reorganization is clearly separated from new interface copy. Source consists of six modules covering paired day ranges and days 11–14. Do not describe a new daily split as 14 original Rita modules or a previously approved schedule.

Verification: Compare exact source passages with the rendered lesson text and a source map. Check all exercise/poem instructions and original downloadable wording.

Evidence and screenshot brief · 1 capture group(s)
local_evidence_ready. Local actual-browser checks113 main plus6 edge checks pass, including complete source comparisons, every daily reflection/video/feedback loop, two participants and31 routes at360/390/430. Daily visual production and independent deployed screenshots pending.

Completed worker evidence and its explicit limits

Required screenshot evidence

Capture the state below at readable scale. Luna first gives a blind description, then checks the visible indicators. Judge all required coverage, not one attractive example.

1. Original exercise text versus rendered lesson0 captured · 0 review record(s)
SCREENSHOT PLACEHOLDER · NOT CAPTUREDReplace with genuine postdeployment captures of the screen and state below.This is a capture brief, not proof or a passed check.
Capture
Source passage and the corresponding expanded daily exercise section.
Set up / action
Show full exercise instructions, examples/poems and reflection wording. Label source range and proposed day split outside the customer text.
Viewport
Internal review canvas/document: 1280×900 CSS px, DPR 1. Open customer screens separately at mobile widths. A canvas thumbnail does not replace a readable mobile screenshot.
Required coverage
All original passages mapped across days 01–14. Capture source/render pairs for each module and every differing daily split.

What independent Luna must be able to see/read:

  • Source wording and meaning remain recognizable and complete in the lesson.
  • Original exercise copy is distinct from new navigation/summary wording.
  • Six source modules are not falsely labelled fourteen originally approved lessons.
Visible reasons to reject: A summary replaces a full exercise or poem. New interface wording quietly changes repetitions, steps or meaning.

Evidence fields to fill: screenshot URL(s) · asset/day ID · state · viewport · capture time · deployed screen URL · actual Luna observations · reviewer report.

Also required. A screenshot cannot prove this:
  • Use exact source-to-day mappings and a normalized text comparison, ignoring extraction whitespace only. Account for six original modules and all fourteen proposed day splits. Do not silently rewrite exercises.

☐ Every day has two actual generated infographic alternatives and two actual AI image alternatives, 56 daily images total, relevant to that day’s content.

Verification: Verify all 56 files, receipts, day assignments and their selectable previews. Inspect text and exercise meaning in each infographic.

Evidence and screenshot brief · 1 capture group(s)
local_evidence_ready. Local actual-browser checks113 main plus6 edge checks pass, including complete source comparisons, every daily reflection/video/feedback loop, two participants and31 routes at360/390/430. Daily visual production and independent deployed screenshots pending.

Completed worker evidence and its explicit limits

Required screenshot evidence

Capture the state below at readable scale. Luna first gives a blind description, then checks the visible indicators. Judge all required coverage, not one attractive example.

1. Four real visual choices for every day0 captured · 0 review record(s)
SCREENSHOT PLACEHOLDER · NOT CAPTUREDReplace with genuine postdeployment captures of the screen and state below.This is a capture brief, not proof or a passed check.
Capture
Daily image chooser/gallery with all four alternatives.
Set up / action
Open chooser for each day; expand both infographics to read their actual labels.
Viewport
Customer screen at 390×844 CSS px, DPR 1 for readable content. Add 360×800 and 430×932 for the stated mobile QA coverage. Never capture a desktop customer UI.
Required coverage
14 chooser capture sets, each showing two infographics and two AI images; all 28 infographics inspected at readable full resolution.

What independent Luna must be able to see/read:

  • Four actual images are present, distinctly labelled by type/alternative.
  • Recommended selection is clear and changing it shows the chosen image in the lesson.
  • Visuals relate to the day’s task and infographic wording agrees with the original instruction.
Visible reasons to reject: A placeholder, duplicate file pretending to be another generation, garbled text or unrelated visual is counted.

Evidence fields to fill: screenshot URL(s) · asset/day ID · state · viewport · capture time · deployed screen URL · actual Luna observations · reviewer report.

Also required. A screenshot cannot prove this:
  • Verify 56 genuine generated files and receipts: day01–14 × infographicA/B and imageA/B. Record unique file hashes, dimensions, type, source purpose and actual chooser mapping.

☐ Each day includes meaningful interactive sliders, video selection and playback, saved demo state, clear submission/feedback states and validation. Selected days begin by rewatching yesterday’s video and reading clearly labelled sample personal feedback. The daily source Napló & önértékelés fields are interactive, with save and readback, rather than only static questions.

Verification: Exercise login/demo access, sliders, file selection, video playback, draft/save/submission and previous-day/feedback states on mobile only. Clearly distinguish local demo storage from a real service. Test reflection/journal edit, save, reload and readback.

Evidence and screenshot brief · 1 capture group(s)
local_evidence_ready. Local actual-browser checks113 main plus6 edge checks pass, including complete source comparisons, every daily reflection/video/feedback loop, two participants and31 routes at360/390/430. Daily visual production and independent deployed screenshots pending.

Completed worker evidence and its explicit limits

Required screenshot evidence

Capture the state below at readable scale. Luna first gives a blind description, then checks the visible indicators. Judge all required coverage, not one attractive example.

1. Readable daily task → reflection → demo submission0 captured · 0 review record(s)
SCREENSHOT PLACEHOLDER · NOT CAPTUREDReplace with genuine postdeployment captures of the screen and state below.This is a capture brief, not proof or a passed check.
Capture
Participant entry/overview and each day’s practice, reflection and video sections.
Set up / action
Enter demo mode, open a day, complete meaningful reflection and select/play fixture video, then demo-submit.
Viewport
Customer screen at 390×844 CSS px, DPR 1 for readable content. Add 360×800 and 430×932 for the stated mobile QA coverage. Never capture a desktop customer UI.
Required coverage
All 14 day views plus entry/overview; capture practice, filled form and submitted status. Include planned previous-feedback checkpoints.

What independent Luna must be able to see/read:

  • The participant can tell what to do, record and submit, and what Rita will do next.
  • Source journal prompts are real controls with visible chosen values/text and save status.
  • Local demonstration labels make clear that a real server upload has not occurred.
Visible reasons to reject: Lesson is only a static document. Controls lack meanings, a video thumbnail is fake playback, or demo success claims a real upload.

Evidence fields to fill: screenshot URL(s) · asset/day ID · state · viewport · capture time · deployed screen URL · actual Luna observations · reviewer report.

Also required. A screenshot cannot prove this:
  • Record interaction receipts for entry, each day’s reflection/save/reload, local media preview, submission and feedback readback. A static screenshot cannot establish playable video or persistence.

☐ Deliver 14 distinct, directly addressable day views plus demo entry/login and a course overview. Every day has a stable day ID, title, correct source range, selected/pending/completed/feedback status and a reachable next action. All 14 can be opened for review without waiting real days.

Verification: Compare day-01 through day-14 in the manifest, canvas and running app. Open each direct URL and each overview tile. No duplicate ID, missing day, incorrect range or unreachable screen may pass.

Evidence and screenshot brief · 1 capture group(s)
local_evidence_ready. Local actual-browser checks113 main plus6 edge checks pass, including complete source comparisons, every daily reflection/video/feedback loop, two participants and31 routes at360/390/430. Daily visual production and independent deployed screenshots pending.

Completed worker evidence and its explicit limits

Required screenshot evidence

Capture the state below at readable scale. Luna first gives a blind description, then checks the visible indicators. Judge all required coverage, not one attractive example.

1. All fourteen days are accessible0 captured · 0 review record(s)
SCREENSHOT PLACEHOLDER · NOT CAPTUREDReplace with genuine postdeployment captures of the screen and state below.This is a capture brief, not proof or a passed check.
Capture
Course overview and first viewport of each individual day.
Set up / action
Enter demo; open every overview tile and each direct day URL without waiting a real day.
Viewport
Customer screen at 390×844 CSS px, DPR 1 for readable content. Add 360×800 and 430×932 for the stated mobile QA coverage. Never capture a desktop customer UI.
Required coverage
One readable overview set plus 14 day-header captures. Capture enough overview sections to read every tile.

What independent Luna must be able to see/read:

  • Days 01–14 have distinct stable IDs/titles and correctly ordered progress/status.
  • Each day header identifies the task, source range/context and useful next action.
  • Selected/pending/completed/feedback states are distinguishable.
Visible reasons to reject: Duplicate IDs, missing tiles, misleading locked days, wrong source ranges or a dead-end screen appears.

Evidence fields to fill: screenshot URL(s) · asset/day ID · state · viewport · capture time · deployed screen URL · actual Luna observations · reviewer report.

Also required. A screenshot cannot prove this:
  • Check day01–14 direct routes, IDs, titles and source ranges in the manifest, overview and running app. Actually open every link.

☐ Every day includes its relevant original rationale, full practice instructions, examples/texts/poems, repetition or duration guidance where supplied, source reflection fields, recording instruction and next step. Distinguish core work from optional extras. Do not compress the lesson to an AI summary or silently omit difficult passages.

Verification: Create a source-to-screen map for every original module passage. Compare rendered text with source, normalizing extraction whitespace only. Record any missing-media marker and intentional day split. All meaningful original instructions and downloadable source content must be accounted for.

Evidence and screenshot brief · 1 capture group(s)
local_evidence_ready. Local actual-browser checks113 main plus6 edge checks pass, including complete source comparisons, every daily reflection/video/feedback loop, two participants and31 routes at360/390/430. Daily visual production and independent deployed screenshots pending.

Completed worker evidence and its explicit limits

Required screenshot evidence

Capture the state below at readable scale. Luna first gives a blind description, then checks the visible indicators. Judge all required coverage, not one attractive example.

1. Complete lesson, not a compressed summary0 captured · 0 review record(s)
SCREENSHOT PLACEHOLDER · NOT CAPTUREDReplace with genuine postdeployment captures of the screen and state below.This is a capture brief, not proof or a passed check.
Capture
All sections of each daily lesson with expandable exercise text opened.
Set up / action
Expand core instructions and any source text/download previews. Capture sections sequentially at readable scale.
Viewport
Customer screen at 390×844 CSS px, DPR 1 for readable content. Add 360×800 and 430×932 for the stated mobile QA coverage. Never capture a desktop customer UI.
Required coverage
All 14 daily content sets and all original module passages. Capture the entire relevant poem/example/instruction, not only its heading.

What independent Luna must be able to see/read:

  • Reason for practice, ordered full instructions, examples, supplied repetitions/duration and original reflection prompts are present.
  • Core exercises and optional extras are clearly separated.
  • Recording instruction and next step complete the lesson; unavailable original media is honestly marked.
Visible reasons to reject: Important material is hidden behind an empty download or AI summary. Poem, repetition instruction or reflection fields disappear.

Evidence fields to fill: screenshot URL(s) · asset/day ID · state · viewport · capture time · deployed screen URL · actual Luna observations · reviewer report.

Also required. A screenshot cannot prove this:
  • Compare all meaningful source passages, rationale, examples/poems, repetition guidance, reflection and downloads to the full rendered text. Keep intentional splitting/missing-media records. Image captures alone cannot establish source completeness.

☐ The day starts with day/progress, a plain task summary and the day’s next useful action. The sequence is previous feedback/rewatch when scheduled, reason for practice, ordered exercises, reflection, recording/submission and what happens next. Exercise steps are readable without a wall of controls.

Verification: Independently review each day as a returning participant. Reviewer must identify what to do first, what to practise/record, how to finish and when feedback appears from the visible screen. Save the concrete confusion found and its corrected screenshot. Do not accept only a DOM/count test.

Evidence and screenshot brief · 1 capture group(s)
local_evidence_ready. Local actual-browser checks113 main plus6 edge checks pass, including complete source comparisons, every daily reflection/video/feedback loop, two participants and31 routes at360/390/430. Daily visual production and independent deployed screenshots pending.

Completed worker evidence and its explicit limits

Required screenshot evidence

Capture the state below at readable scale. Luna first gives a blind description, then checks the visible indicators. Judge all required coverage, not one attractive example.

1. Day header and meaningful task order0 captured · 0 review record(s)
SCREENSHOT PLACEHOLDER · NOT CAPTUREDReplace with genuine postdeployment captures of the screen and state below.This is a capture brief, not proof or a passed check.
Capture
Daily first screen followed by contiguous task/feedback/practice/form/submission sections.
Set up / action
Open each day as a returning participant. On checkpoints open yesterday’s feedback before current practice.
Viewport
Customer screen at 390×844 CSS px, DPR 1 for readable content. Add 360×800 and 430×932 for the stated mobile QA coverage. Never capture a desktop customer UI.
Required coverage
All 14 day first screens and ordered section captures; special checkpoint sequence on days 2,5,8,12 unless a logged source-based alternative replaces it.

What independent Luna must be able to see/read:

  • Day/progress, plain task summary and first useful action are immediately clear.
  • Order is previous review where scheduled, practice reason, ordered exercises, reflection, recording/submission and next expectation.
  • Steps are scannable without controls competing with the instructions.
Visible reasons to reject: The participant must search for what to do first. Recording appears before they understand the task or feedback interrupts unrelated content.

Evidence fields to fill: screenshot URL(s) · asset/day ID · state · viewport · capture time · deployed screen URL · actual Luna observations · reviewer report.

Also required. A screenshot cannot prove this:
  • Independent reviewer must state first action, practice/recording task, completion method and feedback timing from each day. Save actual confusion/revision observations, not guessed participant thoughts.

☐ Convert the source Napló & önértékelés prompts into real frontend controls. Preserve prompts and source scales, show labelled values/endpoints, and keep meaningful free-text reflection. Saving must show a clear status and restore entered answers after reload in the same browser.

Verification: For every day set non-default slider values, enter a distinct reflection, save, navigate away and reload. Read back exact values against that day ID. Empty/unanswered state is distinguishable from a real score. Test failed/unsaved/edited state without server calls.

Evidence and screenshot brief · 1 capture group(s)
local_evidence_ready. Local actual-browser checks113 main plus6 edge checks pass, including complete source comparisons, every daily reflection/video/feedback loop, two participants and31 routes at360/390/430. Daily visual production and independent deployed screenshots pending.

Completed worker evidence and its explicit limits

Required screenshot evidence

Capture the state below at readable scale. Luna first gives a blind description, then checks the visible indicators. Judge all required coverage, not one attractive example.

1. Unanswered versus meaningful filled answers0 captured · 0 review record(s)
SCREENSHOT PLACEHOLDER · NOT CAPTUREDReplace with genuine postdeployment captures of the screen and state below.This is a capture brief, not proof or a passed check.
Capture
Each day’s Napló & önértékelés form.
Set up / action
Capture unanswered state, then change labelled sliders and enter a day-specific reflection.
Viewport
Customer screen at 390×844 CSS px, DPR 1 for readable content. Add 360×800 and 430×932 for the stated mobile QA coverage. Never capture a desktop customer UI.
Required coverage
14 day form sets: unanswered, edited/unsaved, saved and reloaded readback. Include a demonstrated failed-save state.

What independent Luna must be able to see/read:

  • Original prompts and scale labels/endpoints are readable.
  • Unanswered is different from a real zero or middle score; changed values and reflection text are visible.
  • Saved/unsaved/error status is clear and the reloaded form belongs to the same day.
Visible reasons to reject: Default sliders are silently treated as real answers. A save badge appears while displayed values reset or belong to another day.

Evidence fields to fill: screenshot URL(s) · asset/day ID · state · viewport · capture time · deployed screen URL · actual Luna observations · reviewer report.

Also required. A screenshot cannot prove this:
  • For EACH day enter a distinct text and non-default scores, save, navigate away and reload. Compare exact values/day IDs. Test unanswered, edited/unsaved and simulated failed-save states without a backend.

☐ Every day has clear recording guidance, a video chooser, playable preview, replace/remove controls, and a deliberate demo-submit action. A missing or unsupported file cannot appear successfully submitted. Success says it is a prototype and explains that Rita reviews it. Do not imply a server upload.

Verification: Exercise no-file, valid selection, playback, replacement, removal and demo submission on each day using local or bundled fixture media. Confirm the day ID, filename and status shown. If video bytes do not survive reload, keep the saved metadata honest and explicitly request reselection rather than leaving a broken player.

Evidence and screenshot brief · 2 capture group(s)
local_evidence_ready. Local actual-browser checks113 main plus6 edge checks pass, including complete source comparisons, every daily reflection/video/feedback loop, two participants and31 routes at360/390/430. Daily visual production and independent deployed screenshots pending.

Completed worker evidence and its explicit limits

Required screenshot evidence

Capture the state below at readable scale. Luna first gives a blind description, then checks the visible indicators. Judge all required coverage, not one attractive example.

1. No file, valid preview, replace/remove and validation0 captured · 0 review record(s)
SCREENSHOT PLACEHOLDER · NOT CAPTUREDReplace with genuine postdeployment captures of the screen and state below.This is a capture brief, not proof or a passed check.
Capture
Daily recording/video submission area.
Set up / action
Try submission with no file and an unsupported file; then select playable fixture video, play, replace and remove it.
Viewport
Customer screen at 390×844 CSS px, DPR 1 for readable content. Add 360×800 and 430×932 for the stated mobile QA coverage. Never capture a desktop customer UI.
Required coverage
Every day supports these states; capture readable state sets across all 14 IDs. Include actual error and valid-preview captures.

What independent Luna must be able to see/read:

  • Recording guidance, chosen file/day, real video controls, replace/remove and submit actions are visible.
  • No-file/unsupported-file error is understandable and does not show successful submission.
  • Removal restores a clear empty state. Preview shows the intended video rather than an unrelated thumbnail.
Visible reasons to reject: Unsupported media appears submitted, preview is broken, or replace/remove leaves stale filename/status.

Evidence fields to fill: screenshot URL(s) · asset/day ID · state · viewport · capture time · deployed screen URL · actual Luna observations · reviewer report.

2. Honest submission and reload state0 captured · 0 review record(s)
SCREENSHOT PLACEHOLDER · NOT CAPTUREDReplace with genuine postdeployment captures of the screen and state below.This is a capture brief, not proof or a passed check.
Capture
Daily submitted/waiting feedback screen and reloaded video state.
Set up / action
Select valid video and deliberately demo-submit; reload the same day.
Viewport
Customer screen at 390×844 CSS px, DPR 1 for readable content. Add 360×800 and 430×932 for the stated mobile QA coverage. Never capture a desktop customer UI.
Required coverage
Submitted/waiting and reload readback for day01–14.

What independent Luna must be able to see/read:

  • Success explicitly identifies a prototype/local demonstration and explains daily Rita review.
  • Correct day/file metadata remains consistent. If bytes were not retained, reselection is requested and no broken player is shown.
  • Waiting for feedback is different from feedback already received.
Visible reasons to reject: The UI claims an actual server upload or real Rita review. A stale saved filename is rendered as a playable uploaded video after bytes are gone.

Evidence fields to fill: screenshot URL(s) · asset/day ID · state · viewport · capture time · deployed screen URL · actual Luna observations · reviewer report.

Also required. A screenshot cannot prove this:
  • Actually select valid/unsupported/no media, play the preview, replace/remove and demo-submit on every day. Record file/day/status readback and honest reload behavior. Screenshots cannot prove playback or absence of server transfer.

☐ Participant statuses distinguish not submitted, waiting for Rita, feedback available and reviewed. Selected days start with yesterday’s video and specific sample Rita feedback, then relate the next practice focus to the original day’s task. Day 1 has an appropriate start state.

Verification: Use at least days 2, 5, 8 and 12 as planned rewatch checkpoints unless a logged source-based alternative is clearer. Verify previous/current day IDs, playable media and feedback text agree. A demo Rita send must be readable from the matching participant view in the same browser fixture.

Evidence and screenshot brief · 1 capture group(s)
local_evidence_ready. Local actual-browser checks113 main plus6 edge checks pass, including complete source comparisons, every daily reflection/video/feedback loop, two participants and31 routes at360/390/430. Daily visual production and independent deployed screenshots pending.

Completed worker evidence and its explicit limits

Required screenshot evidence

Capture the state below at readable scale. Luna first gives a blind description, then checks the visible indicators. Judge all required coverage, not one attractive example.

1. Waiting, available and reviewed feedback0 captured · 0 review record(s)
SCREENSHOT PLACEHOLDER · NOT CAPTUREDReplace with genuine postdeployment captures of the screen and state below.This is a capture brief, not proof or a passed check.
Capture
Same participant/day before submission, waiting, feedback available and read.
Set up / action
Use one demo fixture, deliberately send feedback from the matching Rita view and return to participant.
Viewport
Customer screen at 390×844 CSS px, DPR 1 for readable content. Add 360×800 and 430×932 for the stated mobile QA coverage. Never capture a desktop customer UI.
Required coverage
All 14 day feedback states; clearly appropriate Day1 start. Focus readable checkpoint sequences on days2,5,8,12.

What independent Luna must be able to see/read:

  • Not submitted, waiting, feedback available and reviewed are distinct states.
  • Yesterday’s video/day ID and specific sample feedback match the previous task.
  • Next practice focus links to the current source exercise, with sample content labelled.
Visible reasons to reject: Feedback arrives before any submission without being identified as sample/demo. Previous-day title/video/message belongs to a different day or gives generic unrelated advice.

Evidence fields to fill: screenshot URL(s) · asset/day ID · state · viewport · capture time · deployed screen URL · actual Luna observations · reviewer report.

Also required. A screenshot cannot prove this:
  • Run browser-local participant submission → Rita draft/send → matching participant readback. Record IDs, exact sent message and state transitions. Do not claim real messaging.

☐ The four visual alternatives for every day are genuine generated assets: two infographics and two AI images. All are available in a chooser, with the recommended option clear. Infographic labels accurately support the source task, have legible Hungarian text and do not introduce contradictory steps or fabricated anatomy.

Verification: Record 56 generated-file identities with dimensions, day IDs, kind and source purpose. Inspect every final image and its mobile rendering. Reject corrupted/gibberish text, stretched images, irrelevant stock-like content or an image replacing essential instructions. Overlapping content across alternatives is allowed.

Evidence and screenshot brief · 1 capture group(s)
local_evidence_ready. Local actual-browser checks113 main plus6 edge checks pass, including complete source comparisons, every daily reflection/video/feedback loop, two participants and31 routes at360/390/430. Daily visual production and independent deployed screenshots pending.

Completed worker evidence and its explicit limits

Required screenshot evidence

Capture the state below at readable scale. Luna first gives a blind description, then checks the visible indicators. Judge all required coverage, not one attractive example.

1. Readable infographic and image purpose0 captured · 0 review record(s)
SCREENSHOT PLACEHOLDER · NOT CAPTUREDReplace with genuine postdeployment captures of the screen and state below.This is a capture brief, not proof or a passed check.
Capture
Day-specific chooser plus expanded infographic and selected in-lesson visual.
Set up / action
Show source task heading nearby; expand image when labels would otherwise be too small.
Viewport
Customer screen at 390×844 CSS px, DPR 1 for readable content. Add 360×800 and 430×932 for the stated mobile QA coverage. Never capture a desktop customer UI.
Required coverage
All 14 days × four alternatives. Inspect every infographic at text-readable scale, plus every selected mobile placement.

What independent Luna must be able to see/read:

  • Infographic labels support the day’s exact exercise, with correct Hungarian and readable layout.
  • AI image has a clear task-related purpose, no fabricated anatomical assertion.
  • Essential exercise wording remains textual and does not depend on reading tiny image captions.
Visible reasons to reject: Gibberish text, unsupported exercise step, cropped label or irrelevant stock-like scene remains.

Evidence fields to fill: screenshot URL(s) · asset/day ID · state · viewport · capture time · deployed screen URL · actual Luna observations · reviewer report.

Also required. A screenshot cannot prove this:
  • Read every infographic against its source task, inspect all 56 original files and final mobile renderings. Verify each chosen image actually changes. Overlap between alternatives is allowed; false anatomy or changed instructions are not.

☐ At 360, 390 and 430 CSS-pixel widths the customer UI remains readable, has no horizontal overflow or overlapping/sticky controls hiding content, supports long Hungarian text, and has usable labelled touch controls. Navigation and submission retain the day context.

Verification: Capture actual postdeployment mobile day sections and interaction states using the required Luna QA/fix/fresh-QA loop. Check all 14 day views and entry/overview. Record no console/runtime errors or broken critical assets. Do not QA larger customer viewports.

Evidence and screenshot brief · 1 capture group(s)
local_evidence_ready. Local actual-browser checks113 main plus6 edge checks pass, including complete source comparisons, every daily reflection/video/feedback loop, two participants and31 routes at360/390/430. Daily visual production and independent deployed screenshots pending.

Completed worker evidence and its explicit limits

Required screenshot evidence

Capture the state below at readable scale. Luna first gives a blind description, then checks the visible indicators. Judge all required coverage, not one attractive example.

1. Participant screens at three mobile widths0 captured · 0 review record(s)
SCREENSHOT PLACEHOLDER · NOT CAPTUREDReplace with genuine postdeployment captures of the screen and state below.This is a capture brief, not proof or a passed check.
Capture
Demo entry, overview and all14 participant days: header, long instruction, form, media and primary actions.
Set up / action
Use actual deployment; capture each section/state at requested mobile widths, including long Hungarian content.
Viewport
Customer screen at 390×844 CSS px, DPR 1 for readable content. Add 360×800 and 430×932 for the stated mobile QA coverage. Never capture a desktop customer UI.
Required coverage
Entry/overview and day01–14 at 360×800, 390×844 and 430×932 CSS px. Cover every section and significant state at readable scale.

What independent Luna must be able to see/read:

  • No horizontal scrolling, overlap, clipped text or sticky control hiding the current task.
  • Labels and touch controls are usable and readable, including long reflection/video states.
  • Day context/navigation remain obvious throughout the screen.
Visible reasons to reject: Desktop-only screenshot substitutes for mobile proof. A pretty first screen hides broken lower sections or obscured submit controls.

Evidence fields to fill: screenshot URL(s) · asset/day ID · state · viewport · capture time · deployed screen URL · actual Luna observations · reviewer report.

Also required. A screenshot cannot prove this:
  • Keep actual 360/390/430 coverage inventory, interaction checks, broken-asset/console results and Luna fix/recheck receipts. No larger customer-screen QA.

☐ Day 14 preserves the original final practice/reflection requirements and offers a clear transition to the online consultation: what to prepare, what will be discussed, how to choose a demo slot and what happens afterwards. Do not imply that 14 days cures a condition or completes all speech development.

Verification: Follow day 14 through final video/reflection, consultation preparation, booking demonstration and confirmation. Compare wording to Rita’s source and the marketing promise matrix. No dead end, invented real availability or mismatch with the paid offer may pass.

Evidence and screenshot brief · 1 capture group(s)
local_evidence_ready. Local actual-browser checks113 main plus6 edge checks pass, including complete source comparisons, every daily reflection/video/feedback loop, two participants and31 routes at360/390/430. Daily visual production and independent deployed screenshots pending.

Completed worker evidence and its explicit limits

Required screenshot evidence

Capture the state below at readable scale. Luna first gives a blind description, then checks the visible indicators. Judge all required coverage, not one attractive example.

1. Final practice → useful consultation0 captured · 0 review record(s)
SCREENSHOT PLACEHOLDER · NOT CAPTUREDReplace with genuine postdeployment captures of the screen and state below.This is a capture brief, not proof or a passed check.
Capture
Day14 final task/reflection/video and consultation preparation/confirmation.
Set up / action
Finish the final demo task, then follow the actual next-action control to preparation and sample booking.
Viewport
Customer screen at 390×844 CSS px, DPR 1 for readable content. Add 360×800 and 430×932 for the stated mobile QA coverage. Never capture a desktop customer UI.
Required coverage
Capture each transition and both participant and Rita consultation contexts.

What independent Luna must be able to see/read:

  • Original final practice/reflection remains intact.
  • The next step states what to prepare, what the online consultation discusses and what follows.
  • Booking is visibly sample/demo; the programme is a practical start, not a guaranteed cure or completed development.
Visible reasons to reject: Day14 ends without a clear next action. The call contradicts the paid offer or displays invented real availability.

Evidence fields to fill: screenshot URL(s) · asset/day ID · state · viewport · capture time · deployed screen URL · actual Luna observations · reviewer report.

Also required. A screenshot cannot prove this:
  • Click final-day reflection/video → preparation → sample slot → confirmation. Compare source and promise matrix. Never infer therapeutic completion or real booking from a prototype.
Shared rules and source boundaries above apply. The check state stays pending until evidence exists.

Fourteen mobile Rita video-review screens

Let Rita review each daily video and write feedback, aligned below the participant journey.

AUTHORIZED0 / 6

Current problem: The review package does not show the actual expert workflow and how feedback returns to the participant.

Proposed change: Build 14 coherent mobile frontend expert views in the canvas row below the matching participant days. Use fixture/local submissions, usable video review and a written-feedback editor with visible demo draft/sent/readback states. No backend, real client portal or delivery service.

Visible at: Second canvas row, directly below matching participant days

Checks and constraints

☐ Each of 14 participant days has a matching Rita screen showing that day’s task context, playable sample/local video, written feedback editor and sent/received states.

Verification: Match all 14 day IDs across rows and exercise a complete browser-local frontend participant submission to Rita feedback and participant readback.

Evidence and screenshot brief · 1 capture group(s)
local_evidence_ready. All14 local expert draft/preview/send/readback loops pass, with matching daily task, reflection and playable fixture. Second participant isolation and live same-browser readback pass. Independent deployed visual/content review pending.

Completed worker evidence and its explicit limits

Required screenshot evidence

Capture the state below at readable scale. Luna first gives a blind description, then checks the visible indicators. Judge all required coverage, not one attractive example.

1. Matched participant ↔ Rita ↔ participant states0 captured · 0 review record(s)
SCREENSHOT PLACEHOLDER · NOT CAPTUREDReplace with genuine postdeployment captures of the screen and state below.This is a capture brief, not proof or a passed check.
Capture
Aligned canvas row plus readable participant submission, Rita editor/sent and participant received view.
Set up / action
Use the same sample participant/day and send a distinctive feedback message.
Viewport
Customer screen at 390×844 CSS px, DPR 1 for readable content. Add 360×800 and 430×932 for the stated mobile QA coverage. Never capture a desktop customer UI.
Required coverage
All14 matched day IDs; each has full context screenshots and feedback readback.

What independent Luna must be able to see/read:

  • Expert screen sits below the matching participant day in the canvas.
  • Submission day, participant, video and feedback editor/sent states refer to the same task.
  • Participant readback displays the actual demo message entered in that Rita view.
Visible reasons to reject: Rows align visually but IDs/text differ. Sending in one day changes another day’s feedback or displays a prewritten unrelated message.

Evidence fields to fill: screenshot URL(s) · asset/day ID · state · viewport · capture time · deployed screen URL · actual Luna observations · reviewer report.

Also required. A screenshot cannot prove this:
  • For all14 IDs exercise the complete browser-local submission/edit/send/readback. Match participant, task, video and exact feedback text. No server, email or messaging proof is required.

☐ Guidance gives concrete observations and the next practice focus without diagnoses or invented individual results. Daily feedback is the primary pilot model.

Verification: Read every demo feedback sample and compare it with that day’s task and the agreed high-service pilot.

Evidence and screenshot brief · 1 capture group(s)
local_evidence_ready. All14 local expert draft/preview/send/readback loops pass, with matching daily task, reflection and playable fixture. Second participant isolation and live same-browser readback pass. Independent deployed visual/content review pending.

Completed worker evidence and its explicit limits

Required screenshot evidence

Capture the state below at readable scale. Luna first gives a blind description, then checks the visible indicators. Judge all required coverage, not one attractive example.

1. Natural task-specific personal feedback0 captured · 0 review record(s)
SCREENSHOT PLACEHOLDER · NOT CAPTUREDReplace with genuine postdeployment captures of the screen and state below.This is a capture brief, not proof or a passed check.
Capture
Rita feedback preview and matching participant feedback panel.
Set up / action
Expand full sample text and show day/task context.
Viewport
Customer screen at 390×844 CSS px, DPR 1 for readable content. Add 360×800 and 430×932 for the stated mobile QA coverage. Never capture a desktop customer UI.
Required coverage
All14 full sample messages, not a cropped preview or one reused paragraph.

What independent Luna must be able to see/read:

  • Each message gives a concrete observation and an actionable next practice focus.
  • It uses supportive natural Hungarian and is clearly sample content.
  • Daily personal feedback is the pilot model without unsupported diagnosis or fabricated personal improvement.
Visible reasons to reject: Generic encouragement replaces an observation/action. Feedback claims an unseen result or diagnostic certainty.

Evidence fields to fill: screenshot URL(s) · asset/day ID · state · viewport · capture time · deployed screen URL · actual Luna observations · reviewer report.

Also required. A screenshot cannot prove this:
  • Read all14 demo feedback texts against the original exercises and daily-feedback pilot proposal. Independent copy reviewer must flag unsupported results/diagnoses.

☐ Each expert view clearly identifies sample participant, day, task/source, submission time/status and the video being reviewed. Switching participant or day changes all dependent content together. Demo data is clearly labelled.

Verification: For all 14 IDs compare the student screen, submission fixture and expert context. Intentionally switch days and verify there is no stale title, reflection, video or feedback belonging to another day.

Evidence and screenshot brief · 1 capture group(s)
local_evidence_ready. All14 local expert draft/preview/send/readback loops pass, with matching daily task, reflection and playable fixture. Second participant isolation and live same-browser readback pass. Independent deployed visual/content review pending.

Completed worker evidence and its explicit limits

Required screenshot evidence

Capture the state below at readable scale. Luna first gives a blind description, then checks the visible indicators. Judge all required coverage, not one attractive example.

1. Unambiguous review context after switching0 captured · 0 review record(s)
SCREENSHOT PLACEHOLDER · NOT CAPTUREDReplace with genuine postdeployment captures of the screen and state below.This is a capture brief, not proof or a passed check.
Capture
Rita queue/review header, task context, video and reflection.
Set up / action
Capture before/after switching between distinct days and sample participants.
Viewport
Customer screen at 390×844 CSS px, DPR 1 for readable content. Add 360×800 and 430×932 for the stated mobile QA coverage. Never capture a desktop customer UI.
Required coverage
Every14 day IDs; include one deliberate day/participant switch comparison.

What independent Luna must be able to see/read:

  • Sample participant, day, task/source, submission time/status and video identity are clear.
  • Video, reflection and feedback context change together with the selected day.
  • Demo data is labelled, with no claim of a real client upload.
Visible reasons to reject: Old title, reflection, video or feedback remains after the selected day changes. Rita cannot tell whose submission is being reviewed.

Evidence fields to fill: screenshot URL(s) · asset/day ID · state · viewport · capture time · deployed screen URL · actual Luna observations · reviewer report.

Also required. A screenshot cannot prove this:
  • Switch day/participant intentionally and compare all dependent content with fixture/source records across day01–14. A screenshot alone cannot prove stale-state prevention.

☐ Rita can open a video, play/pause/replay, read that day’s reflection, write feedback, save/edit a local draft, preview it and deliberately mark it sent in the prototype. Empty feedback cannot be marked sent, and the confirmation identifies recipient/day.

Verification: Exercise the complete frontend workflow with a non-default feedback message on all 14 day IDs. Verify draft restore, editing, empty validation, send confirmation and matching participant readback in the same browser demo. No email/API/backend implementation is necessary.

Evidence and screenshot brief · 1 capture group(s)
local_evidence_ready. All14 local expert draft/preview/send/readback loops pass, with matching daily task, reflection and playable fixture. Second participant isolation and live same-browser readback pass. Independent deployed visual/content review pending.

Completed worker evidence and its explicit limits

Required screenshot evidence

Capture the state below at readable scale. Luna first gives a blind description, then checks the visible indicators. Judge all required coverage, not one attractive example.

1. Review, draft, preview and deliberate demo send0 captured · 0 review record(s)
SCREENSHOT PLACEHOLDER · NOT CAPTUREDReplace with genuine postdeployment captures of the screen and state below.This is a capture brief, not proof or a passed check.
Capture
Rita video/reflect/editor panel, empty-validation, saved-draft and send-confirmation states.
Set up / action
Read reflection, enter a distinctive message, save/reload/edit, preview, then mark sent. Attempt empty feedback separately.
Viewport
Customer screen at 390×844 CSS px, DPR 1 for readable content. Add 360×800 and 430×932 for the stated mobile QA coverage. Never capture a desktop customer UI.
Required coverage
All14 day workflows. Include visible empty-error, restored-draft, preview, recipient/day confirmation and participant readback.

What independent Luna must be able to see/read:

  • Player controls, source task/reflection and editor are readable together or through clear navigation.
  • Draft status and full message are visible; empty feedback cannot appear sent.
  • Preview/confirmation states identify recipient/day and mark delivery as a prototype action.
Visible reasons to reject: Send fires unintentionally, empty feedback passes, or confirmation hides recipient/day. Restored draft is silently replaced or appears on a different day.

Evidence fields to fill: screenshot URL(s) · asset/day ID · state · viewport · capture time · deployed screen URL · actual Luna observations · reviewer report.

Also required. A screenshot cannot prove this:
  • Play/pause/replay real fixture media; save/edit/reload a distinct draft; reject empty feedback; preview/send and compare exact student readback for all14IDs. Frontend only.

☐ Provide a day-relevant sample feedback message with a concrete observation, supportive explanation and an actionable next practice focus grounded in the actual exercise. It must sound natural in Hungarian, be clearly sample content and avoid diagnosis, fabricated personal results or a generic identical paragraph every day.

Verification: Read all 14 feedback samples against the source tasks. Independent reviewer identifies the observation and next action in each, flags unsupported claims and checks differentiation. Record final text and the revision that resolved each material defect.

Evidence and screenshot brief · 1 capture group(s)
local_evidence_ready. All14 local expert draft/preview/send/readback loops pass, with matching daily task, reflection and playable fixture. Second participant isolation and live same-browser readback pass. Independent deployed visual/content review pending.

Completed worker evidence and its explicit limits

Required screenshot evidence

Capture the state below at readable scale. Luna first gives a blind description, then checks the visible indicators. Judge all required coverage, not one attractive example.

1. Fourteen specific, source-grounded feedback examples0 captured · 0 review record(s)
SCREENSHOT PLACEHOLDER · NOT CAPTUREDReplace with genuine postdeployment captures of the screen and state below.This is a capture brief, not proof or a passed check.
Capture
Full daily sample feedback with the matching original exercise context.
Set up / action
Open each sample in the review preview. Do not crop the closing action or its qualifiers.
Viewport
Customer screen at 390×844 CSS px, DPR 1 for readable content. Add 360×800 and 430×932 for the stated mobile QA coverage. Never capture a desktop customer UI.
Required coverage
One readable full-text/context capture per day01–14.

What independent Luna must be able to see/read:

  • Observation, supportive explanation and next focus are identifiable and relevant to this day.
  • The fourteen messages are meaningfully differentiated, natural Hungarian and explicitly sample text.
  • The next action respects the actual exercise instead of adding a new unsupported technique.
Visible reasons to reject: The same generic message is repeated fourteen times. Invented diagnoses, personal results or contradictory technique appears.

Evidence fields to fill: screenshot URL(s) · asset/day ID · state · viewport · capture time · deployed screen URL · actual Luna observations · reviewer report.

Also required. A screenshot cannot prove this:
  • Independent review reads all14 full texts, identifies observation/explanation/next action, checks source fit and logs material before/after copy corrections.

☐ Expert views are mobile only, share the Komplex UI system and expose unread/waiting/draft/sent states clearly. The editor and primary action remain usable with long text, without hiding task/video context.

Verification: Run actual 360/390/430 postdeployment section screenshots and the Luna fix/fresh-QA loop across the 14 expert views and key editor states. Verify readable fields, correct aligned canvas links and no broken media or overflow.

Evidence and screenshot brief · 1 capture group(s)
local_evidence_ready. All14 local expert draft/preview/send/readback loops pass, with matching daily task, reflection and playable fixture. Second participant isolation and live same-browser readback pass. Independent deployed visual/content review pending.

Completed worker evidence and its explicit limits

Required screenshot evidence

Capture the state below at readable scale. Luna first gives a blind description, then checks the visible indicators. Judge all required coverage, not one attractive example.

1. Rita mobile queue, player and long editor0 captured · 0 review record(s)
SCREENSHOT PLACEHOLDER · NOT CAPTUREDReplace with genuine postdeployment captures of the screen and state below.This is a capture brief, not proof or a passed check.
Capture
All14 expert day views in waiting/unread/draft/sent states.
Set up / action
Use actual deployed expert prototype and enter a long Hungarian message; inspect task/video/reflection and primary action.
Viewport
Customer screen at 390×844 CSS px, DPR 1 for readable content. Add 360×800 and 430×932 for the stated mobile QA coverage. Never capture a desktop customer UI.
Required coverage
All14 expert views at 360×800,390×844,430×932. Capture long editor, validation and status states, not only the header.

What independent Luna must be able to see/read:

  • Komplex UI system is consistent and queue/status labels are clear.
  • Long text fields and controls remain readable without overflow or hidden task context.
  • Primary action is reachable and does not cover the player/editor or confuse save with send.
Visible reasons to reject: Desktop QA replaces mobile views. Sticky action obscures content, long feedback clips, or status meanings are unclear.

Evidence fields to fill: screenshot URL(s) · asset/day ID · state · viewport · capture time · deployed screen URL · actual Luna observations · reviewer report.

Also required. A screenshot cannot prove this:
  • Actual 360/390/430 screenshots, independent Luna findings, fixes and fresh deployed recheck across all expert views. Interaction evidence remains separate.
Shared rules and source boundaries above apply. The check state stays pending until evidence exists.

Stripe, emails and final consultation

Complete the connected journey before and after payment.

AUTHORIZED0 / 7

Current problem: Disconnected payment and email previews make it difficult to understand what the buyer receives next.

Proposed change: Build connected frontend-only transaction and email previews: free-video receipt/feedback/nurture, challenge CTA, Stripe-like demo checkout states, confirmation/access, daily delivery and feedback, and final consultation preparation/booking/confirmation/completion. Preserve substantive A/B email coverage. Do not implement Stripe integration or sending.

Visible at: Canvas transaction/email screens and final consultation area

Checks and constraints

☐ Main paid CTA shows the chosen offer, proposed one-time price, what opens next, and Stripe demo checkout. Success, cancellation, pending/error and sold-out/waitlist states are clear. Successful demo purchase opens confirmation/access rather than triggering a real charge. All success/access states are simulated, proving neither payment nor fulfillment. Real payment-event validation and fulfillment belong to the later infrastructure execution.

Verification: Follow each deployed checkout state and verify offer/amount/expectations across landing, checkout and confirmation.

Evidence and screenshot brief · 1 capture group(s)
local_evidence_ready. Root funnel/canvas eight interaction groups pass locally.58 complete email topics and116 A/B texts retain all41 legacy topic IDs and add14 daily personal-feedback and3 consultation-state pairs. Hosted/source/visual acceptance pending.

Completed worker evidence and its explicit limits

Required screenshot evidence

Capture the state below at readable scale. Luna first gives a blind description, then checks the visible indicators. Judge all required coverage, not one attractive example.

1. Truthful frontend payment journey0 captured · 0 review record(s)
SCREENSHOT PLACEHOLDER · NOT CAPTUREDReplace with genuine postdeployment captures of the screen and state below.This is a capture brief, not proof or a passed check.
Capture
Paid landing CTA, Stripe-like checkout, success/access, cancel, pending/error and sold-out/waitlist.
Set up / action
Choose the review offer; navigate each demo state using explicit prototype controls.
Viewport
Customer screen at 390×844 CSS px, DPR 1 for readable content. Add 360×800 and 430×932 for the stated mobile QA coverage. Never capture a desktop customer UI.
Required coverage
All four landing variant destinations as applicable, and every listed checkout branch.

What independent Luna must be able to see/read:

  • Challenge, proposed one-time amount, inclusions and next step agree between landing and checkout.
  • Success opens confirmation/access; cancellation returns to the right offer; pending/error explains sensible retry; sold-out offers a demo waitlist.
  • Prototype/simulated labels are clear and do not imply actual payment or enrollment.
Visible reasons to reject: Checkout changes price/product or celebrates a real charge. Error/cancel/waitlist screens are disconnected or unclear.

Evidence fields to fill: screenshot URL(s) · asset/day ID · state · viewport · capture time · deployed screen URL · actual Luna observations · reviewer report.

Also required. A screenshot cannot prove this:
  • Click deployed checkout branch controls and verify consistent programme/amount/destination. Keep frontend-only fixture/network receipts; never create real Stripe sessions or charges.

☐ All necessary journey emails have full natural copy and matched links, with substantive A/B alternatives. Service messages and optional newsletter/campaign consent are distinct. Paid users are excluded from sales reminders. Preserve the evergreen free-video route, one useful weekly newsletter and a separate fixed-cohort promotional sequence. During a cohort campaign, promotional messages replace the weekly newsletter rather than duplicate it.

Verification: Read every email, inspect its trigger/recipient/next action and verify links, consent and suppression demonstrations.

Evidence and screenshot brief · 1 capture group(s)
local_evidence_ready. Root funnel/canvas eight interaction groups pass locally.58 complete email topics and116 A/B texts retain all41 legacy topic IDs and add14 daily personal-feedback and3 consultation-state pairs. Hosted/source/visual acceptance pending.

Completed worker evidence and its explicit limits

Required screenshot evidence

Capture the state below at readable scale. Luna first gives a blind description, then checks the visible indicators. Judge all required coverage, not one attractive example.

1. Service, opt-in newsletter and cohort sequences0 captured · 0 review record(s)
SCREENSHOT PLACEHOLDER · NOT CAPTUREDReplace with genuine postdeployment captures of the screen and state below.This is a capture brief, not proof or a passed check.
Capture
Canvas email sequence with readable A/B service/nurture/newsletter/promotion previews and consent/suppression cards.
Set up / action
Open complete email bodies, including footer/CTA; inspect subscribed versus not-subscribed and purchased fixture states.
Viewport
Internal review canvas/document: 1280×900 CSS px, DPR 1. Open customer screens separately at mobile widths. A canvas thumbnail does not replace a readable mobile screenshot.
Required coverage
Every email/variant in full inventory and all sequence rules, including weekly newsletter and fixed-cohort campaign.

What independent Luna must be able to see/read:

  • Emails are in the correct recipient/trigger order with complete natural copy and matching links.
  • Required service messages are distinct from optional marketing consent.
  • Promotion replaces the weekly newsletter during cohort campaign and paid participants are shown excluded from sales reminders.
Visible reasons to reject: Uploading a video silently forces marketing subscription. A person depicted as paid keeps receiving purchase reminders or duplicate weekly/promotional messages.

Evidence fields to fill: screenshot URL(s) · asset/day ID · state · viewport · capture time · deployed screen URL · actual Luna observations · reviewer report.

Also required. A screenshot cannot prove this:
  • Read full A/B copy/trigger/recipient/destination inventory and exercise consent/suppression demo fixtures. Ensure campaign replaces rather than duplicates weekly mail and purchasers leave sales reminders. No actual send.

☐ The final day leads to a clear review/next-step online consultation, with preparation, honest demo booking and confirmation states. No unsupported real booking availability or therapeutic completion claims appear.

Verification: Follow the day-14 to consultation path and read the participant and Rita views.

Evidence and screenshot brief · 1 capture group(s)
local_evidence_ready. Root funnel/canvas eight interaction groups pass locally.58 complete email topics and116 A/B texts retain all41 legacy topic IDs and add14 daily personal-feedback and3 consultation-state pairs. Hosted/source/visual acceptance pending.

Completed worker evidence and its explicit limits

Required screenshot evidence

Capture the state below at readable scale. Luna first gives a blind description, then checks the visible indicators. Judge all required coverage, not one attractive example.

1. Clear purpose and next step after the challenge0 captured · 0 review record(s)
SCREENSHOT PLACEHOLDER · NOT CAPTUREDReplace with genuine postdeployment captures of the screen and state below.This is a capture brief, not proof or a passed check.
Capture
Final-day next-action and consultation participant/Rita views.
Set up / action
Open final practice and follow the consultation preparation link.
Viewport
Customer screen at 390×844 CSS px, DPR 1 for readable content. Add 360×800 and 430×932 for the stated mobile QA coverage. Never capture a desktop customer UI.
Required coverage
Capture full preparation and participant/Rita consultation context.

What independent Luna must be able to see/read:

  • The visitor can state when the call happens, what to prepare, what is reviewed and what follows.
  • Sample booking and confirmation are honest; the call reviews practice and useful next steps.
  • No guaranteed cure, completed development or real slot claim is made.
Visible reasons to reject: Consultation is a vague final button without purpose/preparation. Participant and Rita expect different timing or deliverables.

Evidence fields to fill: screenshot URL(s) · asset/day ID · state · viewport · capture time · deployed screen URL · actual Luna observations · reviewer report.

Also required. A screenshot cannot prove this:
  • Walk final-day → consultation on both routes, read participant/Rita views and compare original timing/inclusions. A screenshot cannot prove real booking or therapy completion.

☐ Every paid landing main CTA opens the matching challenge checkout demonstration with consistent offer, proposed amount, inclusions and next action. Success leads to confirmation/access/overview, cancellation returns to the correct offer, pending/error supports a sensible retry, and sold-out leads to a clear waitlist demo.

Verification: Click every landing CTA and all checkout branch controls. Compare offer/version, price and destination at each transition. Test that no real Stripe API/payment session, charge, live waitlist or customer enrollment is created. A set of disconnected screenshots cannot pass.

Evidence and screenshot brief · 1 capture group(s)
local_evidence_ready. Root funnel/canvas eight interaction groups pass locally.58 complete email topics and116 A/B texts retain all41 legacy topic IDs and add14 daily personal-feedback and3 consultation-state pairs. Hosted/source/visual acceptance pending.

Completed worker evidence and its explicit limits

Required screenshot evidence

Capture the state below at readable scale. Luna first gives a blind description, then checks the visible indicators. Judge all required coverage, not one attractive example.

1. Connected checkout, not isolated mockups0 captured · 0 review record(s)
SCREENSHOT PLACEHOLDER · NOT CAPTUREDReplace with genuine postdeployment captures of the screen and state below.This is a capture brief, not proof or a passed check.
Capture
Each paid landing CTA → correct checkout → confirmation/access/overview; cancel/error/waitlist branches.
Set up / action
Click real controls; retain selected offer/version. After cancellation return to the same offer.
Viewport
Customer screen at 390×844 CSS px, DPR 1 for readable content. Add 360×800 and 430×932 for the stated mobile QA coverage. Never capture a desktop customer UI.
Required coverage
Both challenge variants and all critical CTA alternatives; every branch and subsequent access screen.

What independent Luna must be able to see/read:

  • Selected challenge/amount/inclusions stay consistent and the next screen follows from the button label.
  • Success offers clear access/overview, retry retains context, cancellation restores offer, sold-out leads to demo waitlist.
  • Customer is not stranded on a static screenshot with no useful next action.
Visible reasons to reject: A/B variant links open the wrong checkout or price. Success does not lead to access or a branch loses context.

Evidence fields to fill: screenshot URL(s) · asset/day ID · state · viewport · capture time · deployed screen URL · actual Luna observations · reviewer report.

Also required. A screenshot cannot prove this:
  • Exercise every actual main CTA and success/cancel/pending/error/waitlist branch. Record offer/version/price and exact destinations. Confirm no live payment/enrollment/waitlist calls.

☐ Every email preview shows its trigger, recipient stage, subject, preheader, full Hungarian body, CTA destination and substantive A/B alternative. Separate service messages from opt-in newsletter/promotion. Feedback-based follow-up begins after the sample free-video feedback rather than implying Rita already saw an unsubmitted video.

Verification: Review the complete email inventory in journey order, including weekly newsletter and cohort sales messages. Click preview CTAs into the corresponding frontend states. Check purchased participants are not depicted receiving sales reminders and that no placeholder link, arbitrary feature list or unexplained hook survives.

Evidence and screenshot brief · 1 capture group(s)
local_evidence_ready. Root funnel/canvas eight interaction groups pass locally.58 complete email topics and116 A/B texts retain all41 legacy topic IDs and add14 daily personal-feedback and3 consultation-state pairs. Hosted/source/visual acceptance pending.

Completed worker evidence and its explicit limits

Required screenshot evidence

Capture the state below at readable scale. Luna first gives a blind description, then checks the visible indicators. Judge all required coverage, not one attractive example.

1. Full email with trigger and matching next screen0 captured · 0 review record(s)
SCREENSHOT PLACEHOLDER · NOT CAPTUREDReplace with genuine postdeployment captures of the screen and state below.This is a capture brief, not proof or a passed check.
Capture
Mobile email preview showing subject/preheader/body/CTA; A/B comparison and destination.
Set up / action
Expand full body; open the linked frontend state. Place analysis labels outside the customer email.
Viewport
Customer screen at 390×844 CSS px, DPR 1 for readable content. Add 360×800 and 430×932 for the stated mobile QA coverage. Never capture a desktop customer UI.
Required coverage
All emails and substantive A/B alternatives, including newsletter and cohort sales.

What independent Luna must be able to see/read:

  • Trigger and recipient stage are clear in review metadata. Full body, subject, preheader and CTA are readable.
  • Hook builds a natural relevant thought before the offer, not an arbitrary feature list.
  • Feedback follow-up does not pretend an unsubmitted video was reviewed; CTA opens the promised step.
Visible reasons to reject: Placeholder link, unexplained hook, missing subject/body, or sales reminder to a paid fixture survives.

Evidence fields to fill: screenshot URL(s) · asset/day ID · state · viewport · capture time · deployed screen URL · actual Luna observations · reviewer report.

Also required. A screenshot cannot prove this:
  • Read every full email A/B pair against source and journey state, then click every preview CTA. Verify free-video follow-up occurs after feedback and no actual send is performed.

☐ The consultation frontend includes preparation, a clearly sample slot choice, confirmation, an honest meeting-link state, rescheduling/cancellation and a post-consultation next-step view. The source timing of early reservation versus post-challenge consultation is explained consistently.

Verification: Walk both acquisition paths through day 14 and consultation. Reviewer can state when the call happens, what to bring, what the call is for and what follows. Do not invent real availability or a real Meet destination. No booking backend is required.

Evidence and screenshot brief · 1 capture group(s)
local_evidence_ready. Root funnel/canvas eight interaction groups pass locally.58 complete email topics and116 A/B texts retain all41 legacy topic IDs and add14 daily personal-feedback and3 consultation-state pairs. Hosted/source/visual acceptance pending.

Completed worker evidence and its explicit limits

Required screenshot evidence

Capture the state below at readable scale. Luna first gives a blind description, then checks the visible indicators. Judge all required coverage, not one attractive example.

1. Preparation, sample booking and after-call states0 captured · 0 review record(s)
SCREENSHOT PLACEHOLDER · NOT CAPTUREDReplace with genuine postdeployment captures of the screen and state below.This is a capture brief, not proof or a passed check.
Capture
Consultation UI: preparation, slot choice, confirmation, link state, rescheduling/cancel and next-step.
Set up / action
Choose a sample slot; confirm, reschedule and cancel in demo. Open post-consultation view separately.
Viewport
Customer screen at 390×844 CSS px, DPR 1 for readable content. Add 360×800 and 430×932 for the stated mobile QA coverage. Never capture a desktop customer UI.
Required coverage
Every consultation state, plus source-consistent explanation of early reservation versus post-challenge call.

What independent Luna must be able to see/read:

  • Purpose, required preparation, sample date/time and confirmation next step are understandable.
  • Meeting-link state honestly says demo/unconfirmed and does not invent a real Meet destination.
  • Reschedule/cancel has a clear result; after-call screen describes next options rather than claiming a fabricated outcome.
Visible reasons to reject: Invented real availability/link, contradictory date order or dead-end booking state appears.

Evidence fields to fill: screenshot URL(s) · asset/day ID · state · viewport · capture time · deployed screen URL · actual Luna observations · reviewer report.

Also required. A screenshot cannot prove this:
  • Click preparation, sample slot, confirmation, meeting-link state, reschedule/cancel and post-call views. Compare booking timing to sources. No backend calendar or real meeting URL required.

☐ Across ads, landings, checkout, emails, daily participant/Rita UI and consultation, the same service is offered: 14 days of tasks, daily video submission and daily personal feedback, plus the sourced final consultation. Separate 14-day practical value from longer-term development.

Verification: Create one cross-asset promise/destination matrix and read the rendered surfaces against it. Independent reviewer flags contradictions, missing inclusions, unearned guarantees, fake urgency and mismatched CTAs. Fix each material issue in the actual assets before accepting this criterion.

Evidence and screenshot brief · 1 capture group(s)
local_evidence_ready. Root funnel/canvas eight interaction groups pass locally.58 complete email topics and116 A/B texts retain all41 legacy topic IDs and add14 daily personal-feedback and3 consultation-state pairs. Hosted/source/visual acceptance pending.

Completed worker evidence and its explicit limits

Required screenshot evidence

Capture the state below at readable scale. Luna first gives a blind description, then checks the visible indicators. Judge all required coverage, not one attractive example.

1. Same service across the whole journey0 captured · 0 review record(s)
SCREENSHOT PLACEHOLDER · NOT CAPTUREDReplace with genuine postdeployment captures of the screen and state below.This is a capture brief, not proof or a passed check.
Capture
Readable relevant excerpts from ad, landing, checkout, email, daily lesson, expert feedback and consultation beside promise matrix.
Set up / action
Show inclusions and expectation-setting passages in the actual deployed assets, not a newly written summary only.
Viewport
Internal review canvas/document: 1280×900 CSS px, DPR 1. Open customer screens separately at mobile widths. A canvas thumbnail does not replace a readable mobile screenshot.
Required coverage
Every asset variant mapped; include daily tasks/video/daily feedback/final consultation and date/price/limit claims.

What independent Luna must be able to see/read:

  • All stages offer the same fourteen-day practice with daily submissions/personal feedback and sourced final consultation.
  • Proposed price/capacity and unconfirmed dates remain consistently labelled, with no fabricated countdown or urgency.
  • Two-week practical progress is distinct from longer-term development.
Visible reasons to reject: One page reuses old no-feedback promises, another promises a cure, or CTA destinations change the offer.

Evidence fields to fill: screenshot URL(s) · asset/day ID · state · viewport · capture time · deployed screen URL · actual Luna observations · reviewer report.

Also required. A screenshot cannot prove this:
  • Keep cross-asset promise/destination matrix and independent full-copy review. Screenshots cannot prove every claim/source or conversions. Fix all contradictory inclusions, guarantees, fake urgency and mismatched links.
Shared rules and source boundaries above apply. The check state stays pending until evidence exists.

Fast morning review and Rita email draft

Make it easy to choose, comment once and show Rita the right materials.

AUTHORIZED0 / 3

Current problem: A large batch can be complete while still leaving Matt unsure what to show the client.

Proposed change: Provide recommended selections, a brief change guide, unresolved consequential decisions, a complete link index, a concise unsent Rita email and separate acquisition/infrastructure handoff briefs.

Visible at: Canvas introduction and handoff section

Checks and constraints

☐ Every reviewable asset and alternative is linked in the canvas/index. Recommendations are identified as recommendations, with no claim of measured winners or client approval.

Verification: Compare the full artifact manifest to the index and check links on the hosted build.

Evidence and screenshot brief · 1 capture group(s)
pending. Evidence pending. Do not mark passed until the actual required output and verification exist.
Required screenshot evidence

Capture the state below at readable scale. Luna first gives a blind description, then checks the visible indicators. Judge all required coverage, not one attractive example.

1. Complete morning review index and recommendations0 captured · 0 review record(s)
SCREENSHOT PLACEHOLDER · NOT CAPTUREDReplace with genuine postdeployment captures of the screen and state below.This is a capture brief, not proof or a passed check.
Capture
Canvas introduction, recommended path, full categorized index and variant comparison links.
Set up / action
Expand every index group and recommendation/decision panel.
Viewport
Internal review canvas/document: 1280×900 CSS px, DPR 1. Open customer screens separately at mobile widths. A canvas thumbnail does not replace a readable mobile screenshot.
Required coverage
Cover all nine deliverable groups and every manifest asset/alternative with readable index entries.

What independent Luna must be able to see/read:

  • Recommended path is easy to find with substantive alternatives one click away.
  • Day01–14 and matching expert views, all20 ads, all4 landings, email variants, critique, decisions and client email are represented.
  • Recommendations are not labelled measured winners or Rita-approved assets.
Visible reasons to reject: The reviewer must infer filenames or URLs. An image/variant/day/email group is absent or recommendations are falsely presented as proven results.

Evidence fields to fill: screenshot URL(s) · asset/day ID · state · viewport · capture time · deployed screen URL · actual Luna observations · reviewer report.

Also required. A screenshot cannot prove this:
  • Compare complete manifest to every index/canvas link, including all variants/images/days/emails. Verify hosted links. Screenshot of the index alone cannot prove completeness or HTTP success.

☐ A complete unsent Rita email explains each verified URL, what to review and the guiding principles, without asserting unfinished work is done.

Verification: Read the final email against the actual delivered build and follow every URL.

Evidence and screenshot brief · 1 capture group(s)
pending. Evidence pending. Do not mark passed until the actual required output and verification exist.
Required screenshot evidence

Capture the state below at readable scale. Luna first gives a blind description, then checks the visible indicators. Judge all required coverage, not one attractive example.

1. Unsent Rita email with precise review instructions0 captured · 0 review record(s)
SCREENSHOT PLACEHOLDER · NOT CAPTUREDReplace with genuine postdeployment captures of the screen and state below.This is a capture brief, not proof or a passed check.
Capture
Final Hungarian email draft on the review surface.
Set up / action
Show subject and complete body, including all URL descriptions and review questions.
Viewport
Internal review canvas/document: 1280×900 CSS px, DPR 1. Open customer screens separately at mobile widths. A canvas thumbnail does not replace a readable mobile screenshot.
Required coverage
Capture the entire email at readable scale; follow and independently verify every included URL.

What independent Luna must be able to see/read:

  • Email tells Rita what each URL contains, what to review and the guiding principles.
  • It describes the actual delivered frontend review/proposals honestly and explains consequential decisions needing approval.
  • Draft/unsent status is explicit in review metadata, with no claim of sending or client approval.
Visible reasons to reject: Generic email lists unnamed links, claims unfinished assets are finished, or makes unapproved binding promises.

Evidence fields to fill: screenshot URL(s) · asset/day ID · state · viewport · capture time · deployed screen URL · actual Luna observations · reviewer report.

Also required. A screenshot cannot prove this:
  • Read the UNSENT client email against actual finished assets and verified URLs. Follow every link and ensure unfinished/approval-needed items remain honest. No send is authorized.

☐ At least two independent real Luna xhigh QA partitions, Luna max defect fixes, and fresh postdeployment mobile screenshots cover all changed client pages/sections and significant interaction states.

Verification: Inspect launch receipts, coverage inventory, section/run screenshots and final defect state. Do not test larger customer-interface screen sizes.

Evidence and screenshot brief · 2 capture group(s)
pending. Evidence pending. Do not mark passed until the actual required output and verification exist.
Required screenshot evidence

Capture the state below at readable scale. Luna first gives a blind description, then checks the visible indicators. Judge all required coverage, not one attractive example.

1. Independent observations and complete capture index0 captured · 0 review record(s)
SCREENSHOT PLACEHOLDER · NOT CAPTUREDReplace with genuine postdeployment captures of the screen and state below.This is a capture brief, not proof or a passed check.
Capture
QA evidence index linking actual mobile screenshots and blind Luna reports.
Set up / action
Show asset/section/state/viewport, deployed build and review/fix/recheck references. Open readable samples.
Viewport
Internal review canvas/document: 1280×900 CSS px, DPR 1. Open customer screens separately at mobile widths. A canvas thumbnail does not replace a readable mobile screenshot.
Required coverage
Every changed customer page/section and significant state at 360/390/430, split into at least two independent QA partitions.

What independent Luna must be able to see/read:

  • Each evidence item identifies the exact deployed asset, state and viewport, with actual screenshot and independent observations.
  • Original defect, corrective action and fresh recheck are linked without hiding outstanding issues.
  • Only mobile customer captures are counted. Review canvas screenshots remain a separate document category.
Visible reasons to reject: A green count has no readable underlying evidence. Reports repeat expected rubric as if observed or use local/stale/desktop customer captures.

Evidence fields to fill: screenshot URL(s) · asset/day ID · state · viewport · capture time · deployed screen URL · actual Luna observations · reviewer report.

2. Same defect before and after on the deployment0 captured · 0 review record(s)
SCREENSHOT PLACEHOLDER · NOT CAPTUREDReplace with genuine postdeployment captures of the screen and state below.This is a capture brief, not proof or a passed check.
Capture
Readable before/after mobile section screenshots for every material defect.
Set up / action
Capture corrected section/state at same viewport on the final deployed build; get fresh independent observations.
Viewport
Customer screen at 390×844 CSS px, DPR 1 for readable content. Add 360×800 and 430×932 for the stated mobile QA coverage. Never capture a desktop customer UI.
Required coverage
All significant visual/content/UX defects from both partitions, with matching before/after contexts.

What independent Luna must be able to see/read:

  • The specific defect is visible before and resolved after, at the same readable scale.
  • Nearby content remains usable and the final capture genuinely belongs to the deployed revision.
Visible reasons to reject: The evidence only asserts fixed or changes viewport/crop to conceal the issue.

Evidence fields to fill: screenshot URL(s) · asset/day ID · state · viewport · capture time · deployed screen URL · actual Luna observations · reviewer report.

Also required. A screenshot cannot prove this:
  • Keep at least two independent Luna xhigh partition launch/report receipts, Luna max fixes, coverage inventory and fresh deployed re-QA. Report actual findings; no screenshot-only greenwashing.
Shared rules and source boundaries above apply. The check state stays pending until evidence exists.

Settled defaults and remaining source gaps

Do not block frontend execution on demo data

Matt authorized execution. Use the proposed defaults unless the clarified frontend scope or a logged rational amendment changes them. Unconfirmed dates, recordings and booking availability stay explicitly demo/proposed. Do not ask the sleeping user to approve settled scope again.

Q1 · Is the complete deliverable package right?

Proposed answer: Keep all nine work packages: outer critiques, meta-persona and step reasoning, journey canvas, 20 new ads, four landings, 14 participant days, 14 Rita screens, checkout/emails/consultation, and the morning/client handoff. Generate all 83 requested visuals. Customer screens and their QA stay mobile only.

Your input: Name any missing deliverable or scope change. Everything else stays included.

Q2 · How should all 28 template layouts be used?

Proposed answer: The current strict interpretation uses S1–S28 meaningfully on each of the four landing pages. Keep two variants per offer and three complete main CTA-section alternatives per offer. Preserve Komplex identity and use Gemini 3.1 Pro for the four landing generations.

Your input: Confirm all 28 on every page, or choose coverage across the four-page family instead. This affects length substantially. No repetitive filler or fabricated proof is acceptable either way.

Q3 · Is 10 free-video ads plus 10 direct-challenge ads the right split?

Proposed answer: Yes. Twenty distinct 1:1 concepts, no Rita name, with matching Facebook text and a clear destination. Luna critiques each proposed caption and its meaning before generation. The canvas identifies hypotheses and recommended starting tests. Begin with proposed broad Meta Advantage+ targeting. Put the settings and testing assumptions in the review, without activating campaigns.

Your input: Change the split or flag an angle you want included. Separate audience segments are optional test angles rather than a persona-research project.

Q4 · Which pilot terms should the review propose to Rita?

Proposed answer: 24,900 Ft once, 12 participants, daily task guides, daily participant videos and daily personal Rita feedback. Free-video feedback within three working days. These remain Matt’s proposals for Rita, not claims of existing Rita approval.

Your input: Keep these terms or replace the amount/cap/service detail. Four feedback batches stay only a future efficiency alternative.

Q5 · What dates and daily service cutoffs should be shown?

Proposed answer: No invented campaign dates or resetting urgency. Show clearly labelled demonstration dates until the real opening, signup closing and start dates are agreed. A visible service-timing proposal can use a daily 18:00 video cutoff and feedback by 18:00 the next day, including an explicit weekend rule for Rita to review.

Your input: Give real campaign dates if available. Choose the daily submission/response times and weekend handling, or leave the timing proposal clearly unapproved in the review.

Q6 · Are Rita’s real instruction video and consultation-booking link available?

Proposed answer: Use verified media/booking destinations if supplied. Otherwise show honest demo states. Keep browser-local prototype data, demo Stripe states and unsent email copy. Real accounts, hosted uploads, charges, fulfillment and sending belong to a later infrastructure loop.

Your input: Provide the source/video and booking links, or keep demo placeholders for this review. Optional source gaps do not block completing the review set.

Working assumptions and source gaps

Working assumption · interactive review prototype

The web app saves demonstration state locally, including a local video preview. Stripe and email states are simulated. A later infrastructure loop implements real accounts, video storage, payment fulfillment and sending.

Working assumption · variant coverage

Keep two landing variants for each offer, three critical CTA alternatives, substantive email A/B coverage, and a proposed 10 free-video + 10 direct-challenge ad split. All28 kit layouts appear meaningfully across the four-page family under the logged rational amendment. All four pages remain complete and materially different, with50 purposeful sections instead of112 repetitive blocks.

Open input · pilot terms

24,900 Ft once and 12 participants carry forward as review proposals, not a binding new client commitment. Daily feedback remains the core pilot service. The response timetable and workload assumptions must be plainly visible to Rita.

Open input · real dates

Starting date and signup closing date are not fixed. A countdown cannot imply a real deadline until it is agreed. Demo dates are visibly marked.

Open input · Rita video and real booking URL

The source exercise text is available. Rita’s recorded instruction/welcome video and the actual final-consultation booking destination still need a verified source. The review labels teaching videos as unavailable and booking as a local demo. The technical playback fixture is neither a Rita lesson nor a participant recording.