Does the AURORA prototype deliver every Phase 1, 2 and 3 capability in the decision sheet?

Yes, after remediation. All 114 Phase 1–3 capabilities are now demonstrable in the prototype, each in the build its phase belongs to, and the 16 Phase 4+ research questions are stated rather than claimed. The uploaded prototype passed 52 of 114; the other 62 were thin, mis-phased or missing and have been built. A further 9 Phase 4+ items lacked an honest statement and are now stated.

Phase 1 · MVP

58/58

before: 44 of 58 pass

Phase 2

32/32

before: 7 of 32 pass

Phase 3

24/24

before: 1 of 24 pass

Phase 4+ · stated as research

16/16

before: 7 of 16 stated

How the assurance ran

1 · IngestEvery row of all five sheet tabs (D1–D8, 58-item build, 130-row roadmap). Rebuilt the missing runtime so the three apps run offline.

2 · FrameworkVerdict scale per capability, phase-correct evidence only; roadmap and decision lists never count as evidence.

3 · Live auditEvery screen, module and tab rendered in each build (111 views × 3 apps). Five auditors graded all 130 items and found donors in Client and Internal.

4 · RemediateOne capability-panel grammar for every gap, plus structural fixes: chips, tenant, roles, period, entity, engines (D4).

5 · VerifyAutomated: 47/47 panels appear only in their phase build; 17/17 interaction checks; 0 runtime errors.

Decision conformance (D1–D8)

#	Decision	Status	Evidence
D1	Phase 1 scope: 58 capabilities	Conforms	All 58 demonstrated in the Phase 1 build; traceability below.
D2	Build order, foundations first	Conforms	Phase 1 decisions screen shows waves 0 → 3 + Commercial with lock order.
D3	Price on assured measurement; unlimited seats; not $89	Conforms	Plan ladder prices on assurance tiers; final number waits for the cost model (owner action).
D4	4 engines at launch; add without rebuild	Fixed → conforms	Uploaded prototype shipped 8 at launch. Now 4 (P1) → 6 (P2) → 8 (P3), each a configuration with a break marker.
D5	Measure independently; HITL; nothing auto-publishes	Conforms	Read-only at launch; Phase 3 HIVE → brand/legal review → SYNDICATE handoff with approval gates and injection screening.
D6	Defer sentiment, traffic link, attribution, deep competitor tools	Conforms	Tone only with published rubric + κ (P2); pre-registered protocol (P2); attribution (P3). Optional intent view: Intent chip live in P1.
D7	Two checks before building	Owner action	Walk-through of the 6 built items recorded in product; the cost model is outside the prototype.
D8	First milestone: assured measurement	Conforms	Build switch shows the Phase 1 assured-measurement milestone first.

Traceability: all 130 capabilities

Q	Phase	Capability	Before	Gap found	Built / fixed, and where	After

Design review (19 findings)

#	Severity	Finding	Status	What changed
DS-01	high	Tablet portrait header overflows the 820px artboard: account button 'CG Chris Green' is pushed 100px past the right edge and clipped by the frame's overflow:hidden (only the 'C' of the avatar is visible). Header content …	Fixed	Search box flexes; name and workspace hidden on tablet
DS-02	high	Tablet navigation is incomplete and mislabels the current screen. The header 'Section' select only has start/snapshot/modules/evidence/methodology/account, so Actions, Reports, BI reports, Discoverability, Attribution, G…	Fixed	Tablet Section menu lists every screen and tracks the current one
DS-03	high	Wide tables are silently clipped: <main> has overflow-x:hidden, so the decision table (821px in a 752px card) and the 130-row roadmap table (919px in 818px) lose their right-hand columns (Rationale, Owner, Build status) …	Fixed	Wide tables sit in labelled sideways-scroll regions
DS-04	high	Reporting period is inconsistent across the product: filter bar says 'Period August 2026' on Snapshot and Modules but 'September 2026' everywhere else; Modules grid cards under an August filter all say '... in September'…	Fixed	One reporting period everywhere; Period chip switches August/September
DS-05	high	Contradictory verdicts on the same figure: the Influence Share card says '39 ... Fell 5 points · September vs August · in line with expected' (39 is inside expected 39-45), while the trend annotation directly below says …	Fixed	Change and expectation read consistently: fell vs August, in line with expected
DS-06	high	Text-on-fill contrast failures inside data cells. Gains heatmap uses white text on rgba(22,48,42,alpha) cells; at alpha 0.42 the effective fill is ~#9EA9A6 (white ~2.4:1) and the secondary '6 prompts' line has opacity .8…	Fixed	Heatmap text colour chosen by fill strength
DS-07	medium	Known issue confirmed. All 10 selects render 288px wide inside a 248px content box and overflow the panel's right edge by 39px (23px past the panel border, 7px off-window at 1464px). Cause: the inner panel <div style="di…	Fixed	Prototype controls: selects fit the panel
DS-08	medium	Collapsed 'Prototype controls' pill (position:fixed bottom-right, z-index 100) permanently sits on top of the artboard and covers live content: action buttons on Actions ('Send to HIVE'/'See the finding' row of card 2), …	Fixed	Compact pill; Shift+P or ?present hides controls for showcases
DS-09	medium	Range strips plot outside their 96px track (no axis, no clamping): e.g. Trial 'Owned 72%' marker is drawn ~44px beyond the strip and past the table edge; table becomes 1100px in a 1092px card.	Fixed	Range strips: percentage scale, clamped to the track
DS-10	medium	Bar chart with a truncated y-axis (starts at 30): August (44) and September (39) bars start below the plot baseline, exaggerating the 5-point fall visually. Breaks the 'restricted chart grammar' (bars must start at zero;…	Fixed	Zoomed axis marked (≈ 0) and stated on the chart
DS-11	medium	Who-holds-the-answers stacked bars use six saturated categorical hues with 48 inline numbers; ranges reduced to a 2px white tick on one series only. Off-grammar vs the rest of the app (ink/green, ranges always visible) a…	Open	Stacked share keeps six brand hues (polish backlog)
DS-12	medium	Interval rule broken on the module grid: 8 of 12 cards show direction words ('Fell 5 points', 'No clear change') with no likely range or freshness; only Brand description, Competitors, Opportunities, Integrity, Journey s…	Fixed	Every module card shows value, likely range and data-to date
DS-13	medium	Onboarding step 1 ('Which brand, and where?') still shows the full sidebar and the populated filter bar (Entity Harbour & Pine, Competitors 5 named, Period September 2026) - the brand being set up already appears to have…	Fixed	Onboarding hides sidebar and filter bar
DS-14	medium	Decision rows compress to four narrow columns at 820px: titles wrap to 4-6 lines, decision note inputs truncate their value ('Four checks to close first;'), and H1 breaks awkwardly ('Phase / 1').	Improved	Decision table scrolls on portrait; row layout unchanged
DS-15	low	Charts overflow their containers by 4-10px (axis labels / end labels past the card padding): competitor dot plot 1099/1092, funnel 1096/1092, attribution charts 1096/1092, gains Summary bars 1096/1092, description bars 5…	Open	4–10 px label overhangs on some charts (polish backlog)
DS-16	low	Ellipsis truncation hides meaning: gains source-of-change row/column labels truncate to 44px at 820 ('Comparative', 'Oakfield Home', 'comparison questions' all cut), measure tree node titles truncated at ipadP, action ti…	Open	Label truncation at 820 px portrait (polish backlog)
DS-17	low	Sub-12px text: funnel SVG labels 'DROPPED'/'REACHED' at 11px, footnote superscripts at 10px.	Fixed	No text below 12 px
DS-18	low	ARIA tab semantics are inconsistent: 14 role=tablist but only 7 role=tabpanel; sort toggles ('Sort by alongside / instead of') and period toggles ('Sep vs Aug / Aug vs Jul') are marked role=tab although they re-sort/re-p…	Open	ARIA tab/toggle semantics tidy-up (polish backlog)
DS-19	low	Snapshot is 4,744px tall at desktop (5.3 screens) with ten sections; the 'What next' section shows no actions inline (only a 'See the three actions' button), so the decision content sits below the fold while the measure …	Open	Snapshot length; sections already user-toggleable

What it looks like

Snapshot, Phase 1: one trustworthy score with range, period-consistent

Icon-grid module home with range + freshness

Filters carried: entity + intent, evidence filtered

Engine adapters and health (Q7, Q30, Q90)

Phase 2: crawler access, validated (Q25)

Phase 2: tone only with published rubric (Q41, Q44)

Phase 2: pre-registered traffic protocol (Q51)

Phase 3: HIVE → review → SYNDICATE handoff

Showcasing it

Open AURORA Prototype.html (works offline, any modern browser). Press Shift+P or add ?present to the URL to hide the reviewer controls.
Set Build in Prototype controls to Phase 1, 2 or 3. Engines, modules and capability panels appear as each phase lands; tags show the phase on every panel.
Walk the six flows on Start here. Use the Period chip: August shows a normal month, September shows a real fall and how AURORA explains it.
The agency side is AURORA Internal.html (linked from both apps). AURORA Client (reference).html is the unmodified donor design.

For the owner

D3 and D7(b): the price point and the cost model are business inputs outside the prototype.
Decisions tab: the amber Decision and sign-off cells are left for you. The assured workbook only adds an Assurance (130) tab.
Sample data: Harbour & Pine figures are illustrative and internally consistent. They are not live measurement.
Polish backlog: five low or medium design items are open (above). None blocks a showcase.

Sources: Aurora_MVP_Decision_Sheet.xlsx, AURORA_Prototype.dc.html, AURORA_Client.dc.html, AURORA_Internal.dc.html. Raw audit data in audit_data/; matrix also in traceability_matrix.csv and the assured workbook.