aplus-module-generator

name: aplus-module-generator description: Generate the 5-7 vertical A+ Premium modules for an Amazon listing — Higgsfield image generation + structured layout following the two-half architecture (top half capture & convert, bottom half visual FAQ). Use when running aplus-module-generator {ASIN} or when planning A+ Premium content that follows Keplo's methodology.

aplus-module-generator — A+ Premium 5-7 Module Set

Per 02-visual-content/a-plus-content.md, A+ Premium uses two-half architecture: top 3 modules = capture & convert with photography-forward emotional content, bottom 4 modules = inform & reassure (visual FAQ).

Methodology — read before rendering

Read reference/02-visual-content/aplus-creative-director.md in full before generating any module. (Updated in v0.1.7 — A+ has its own dedicated creative-director ref now, separate from listing-image.) That file is the source of truth for: the 13-step A+ flow, the 4 A+ Critical Rules (Legibility / Single Message / Visual Benefit / Branding consistency), the 5 Pre-Generation Checklist questions, editorial design philosophy, anti-clipart rules, the 6-section mandatory prompt structure (A+ variant), 4-axis scoring rubric, comparison-module exception, A+ module taxonomy, and iteration loop.

Hard rules specific to A+ modules:

• Aspect ratio: 16:9 (widescreen, horizontal layout) — locked in at start AND end of every prompt. NOT 1:1. Wrong-ratio outputs trigger automatic re-generation.
• Reference image always: pull live product photo from brain/products/{asin}-{geo}.json and pass to Higgsfield
• Brand guidelines respected: brain.business.brand_guidelines drives palette + visual style keywords + type placement
• No text generation in image: render clean photography; brand designer adds type in post
• No clipart elements — no callout arrows, no dimension lines, no overlay boxes, no flow charts, no annotated diagrams, no burst stickers. A+ modules are editorial photography with type, not infographic clipart.

A+ comparison modules requiring side-by-side spec tables follow the editorial-diagram exception in the listing creative-director reference — clean type, generous white space, photo of product as anchor.

Output emits florence-concepts-{asin} per the artifact protocol.

Prerequisites

• Higgsfield MCP
• Research brief (asin-deep-research) and Rufus gap analysis (rufus-gap-analysis)

Invocation

aplus-module-generator {ASIN}
aplus-module-generator {ASIN} --modules=7   # 5-7, default 6
aplus-module-generator {ASIN} --premium     # use full-width hero + video

Output

/tmp/cro-content/{ASIN}-aplus-{date}.md with 6 modules (1 hero, 2 capture, 3+ FAQ) + image set in /tmp/cro-content/{ASIN}-aplus/.

Phase 1 — Source Research

Pull from existing files:

• Top 3 purchase drivers → modules 1-3 (capture)
• Top 3-4 Rufus FAQ-class queries → modules 4-7 (FAQ)
• Photo review demographics → module 1 hero

Phase 2 — Module Plan

Phase 3 — Generate Each Module

For each module:

• Image: Higgsfield call with module-specific prompt. Default Nano Banana Pro. Aspect varies (full-width 970×600 or two-up 487×487).
• Copy: Pull from research brief verbatim where possible. Keep modules scannable in 15-20s total.
• Alt-text: Pack with product name + keywords + benefit + use case + audience (Amazon search + Rufus index this).

Phase 4 — Validate (per 02-visual-content/a-plus-content.md)

• ✅ Top half is photography-forward, not specs
• ✅ Bottom half is visual FAQ (not text-only)
• ✅ No duplication of listing slot 2-7 images
• ✅ Full-width preferred over side-by-side on mobile
• ✅ Total scan time ≤20 seconds
• ✅ Alt-text on every image

Phase 5 — Output

# A+ Premium Module Set — {Title}

**ASIN:** {ASIN} | **Date:** {date} | **Modules:** {N}

## Two-Half Architecture

### Top Half — Capture & Convert
1. **Hero — emotional moment** ![](module-1.png)
   - Copy: {short headline + 1 sentence}
   - Alt-text: {keyword-packed string}
2. **Benefit deep-dive** ![](module-2.png)
   - ...

### Bottom Half — Inform & Reassure
4. **Visual FAQ — {Rufus query}** ![](module-4.png)
   - ...

## Designer Brief

- Image specs: 970×600 full-width (modules 1, 6); 487×487 two-up (others)
- Format: PNG, sRGB, ≤500KB each
- Use Higgsfield-generated as composition reference, polish for retail
- Do NOT duplicate any listing image (slots 1-7) — A+ should add, not repeat

## Recommended Next Skills

- Validate scannability: `aplus-comprehension {ASIN}`
- Build the FAQ alt-text pack: `rufus-answer-pack {ASIN}`
- Convert to A+ Premium spec doc: `anthropic-skills:amazon-aplus-brief`

Reference Files

• Vault: CRO-Knowledge-Base/02-visual-content/a-plus-content.md (two-half architecture)
• ~/.claude/skills/cro/aplus-best-practices.md
• Anthropic skill: amazon-aplus-brief (companion for client-ready Word doc)

Quality Bar

• All 6 (or 7) modules have research-traced source
• Top half ≠ specs/FAQ (capture rule)
• Bottom half = visual FAQ (inform rule)
• No duplication of slot 2-7 images
• Alt-text packed on every image (often skipped — high SEO value)

Auto-Triggers

• User asks "build A+ for {ASIN}"
• /cro-content-plan Phase 3 for A+
• After rufus-gap-analysis to act on the gaps

v0.1.12 — Read brain.image_strategy before generating

Before drafting any Higgsfield prompt, read brain.image_strategy from working memory:

• If null → pause and offer image-strategy first (~10 min, makes every future render bespoke). If user proceeds without, flag concept cards with "No category research used — generic aesthetic."
• If set + fresh (<90d) → inject prompt_adjustments.scene_keywords into prompt section 1, palette_keywords + mood_keywords into section 4, and do NOT include {anti_patterns_csv} near the end. Cite the strategy on each concept card.
• If stale (>90d) → flag and recommend image-strategy --refresh.

This brand-level strategy comes from skills/image-strategy.md's top-15-bestsellers analysis. Same adjustments apply across every render for this brand.

v0.1.13 — Visual verification gate + base64 embedding (NON-NEGOTIABLE)

Florence does NOT present an image she hasn't actually looked at, AND she does NOT use raw Higgsfield URLs in artifact HTML. Two blocking rules added in v0.1.13:

Verification gate

After Higgsfield returns each generated image, BEFORE adding to any artifact OR sending to Pinion:

1. Load the image into Florence's multimodal context — paste the URL in chat so Cowork's multimodal Claude loads it natively. Narrate: "Looking at #{N} before I include it."
2. Visually inspect against: aspect ratio (1:1 main+listing / 16:9 A+), product fidelity vs reference, technique landed (visibly), background appropriate, no clipart leak (anti-clipart rules), no text-on-main-image (TOS), no model faces.
3. If anything fails → re-prompt with specific fix + regenerate. Cap at 3 attempts per concept; surface honestly on attempt 4.
4. Only after ALL concepts pass → proceed to artifact emission.

The eye trumps the score. Even if the rubric said 100/100, if visual inspection finds a wrong product or off-aspect output, it fails the gate.

Base64 embedding

Higgsfield URLs are temporary AND Cowork's artifact iframe sandbox blocks external image loads in many builds. For every verified concept:

1. HTTP GET the Higgsfield URL → fetch image bytes
2. Detect MIME type from response headers (typically image/png)
3. Base64-encode the bytes
4. Substitute {{image-src}} (or {{winner-image-src}} for tests) with data:image/png;base64,<encoded> — NOT the raw Higgsfield URL

This makes the artifact self-contained — survives sandbox + URL expiry. ~1-3 MB per image is fine for Cowork.

The live product image ({{product-image-url}} in concepts.html hero strip / dossier head / cockpit product cards) stays as the live Amazon CDN URL — that's permanent and not affected.