Flova Ecommerce Visual Kit

Ecommerce Visual Kit

Mission

用于电商产品视觉全案：基于产品图生成一套静态电商视觉资产，包括 3 张核心主图和 6-8 张详情页图，并用产品基线审计保证形态、颜色、材质、Logo 和朝向一致。

This workflow is static-image-first. Unless the user explicitly asks for video, do not route to video generation or timeline assembly.

JarvisHub Execution Model

• 主 Agent 负责编排：读取画布事实、整理任务 brief、按阶段同步 TodoWrite 进度，并把已确认的文本成果写入画布；媒体生成、等待、拼接和评审交给具备相应能力的执行 agent 或当前可用工具。TodoWrite 不是画布写入前置条件。
• 图像、视频和拼接交给具备媒体能力的执行 agent：brief 给出稳定输出身份、用途、真实参考 URL 和关键约束；需要下游引用时必须等待真实 imageUrl / videoUrl，不能把提交态当完成态。
• 多模态验收交给 critic sub-agent：只读取真实媒体并评审，不生成、不补素材。
• 当前已知 canvas 工具集没有通用音频生成、音频驱动口型、末帧抽取或视频直改工具；旁白、BGM、字幕、末帧承接和精准 lip sync 只能作为后期合成计划或 blocked 项，除非本轮工具列表明确暴露对应能力。

When To Use

Use this skill for:

• ecommerce product visual kits,
• product hero images,
• product detail page images,
• white-background product KV,
• feature-benefit images,
• material macro images,
• before/after comparison images,
• product size/spec layout,
• static product campaign image set.

Use flova-product-commercial-short if the user wants product video. Use canvas-brand-web-design if the user wants a webpage, not a product image kit.

Canvas-Native Boundary

Follow the JarvisHub Execution Model for product analysis, image generation or editing, review, and gallery output.

If the harness cannot compute exact HSL/logo coordinates, produce a best-effort visual baseline and review checklist. Do not claim numeric precision that was not actually measured.

Required Intake

Required:

• product_reference: uploaded product image.

Useful optional inputs:

• product name,
• target platform,
• target aspect ratios,
• brand colors,
• style direction,
• selling points,
• forbidden colors/props/claims,
• whether Logo/text may be visible,
• whether white-background images are required,
• number of detail-page images.

If no product image exists, ask for it before generation. You may draft a shot list from text, but do not claim product-faithful visuals.

Workflow

1. Analyze product baseline.
2. Create ecommerce product spec.
3. Plan 3 hero images and 6-8 detail images.
4. Bind uploaded product as Product_Asset_01.
5. Generate images in storyboard order.
6. Audit consistency.
7. Return image set and review report.

Pause for confirmation after:

• spec,
• storyboard/image plan,
• first generated hero image or first batch,
• completed kit/audit.

For autonomous runs, use critic review instead of user confirmation and state that decision.

Product Baseline

Extract and store:

• outline and proportions,
• primary/secondary/highlight color,
• material and finish,
• reflectivity/gloss behavior,
• logo/text location and style if present,
• structure zones,
• default facing direction,
• key functional feature,
• image quality issues.

Baseline example:

Product_Asset_01: cylindrical bottle, height:width about 2.4:1, brushed metallic cap, matte body, logo centered upper front, default facing left-oblique around 30 degrees, primary cool blue-gray, soft vertical specular highlight.

Image Kit Structure

Hero Images

Generate three platform main images in this order:

Detail Images

Generate 6-8 detail images in this suggested order:

Do not change this order unless the user or platform needs a different kit.

Shot Fields

Each planned image must include:

• Shot ID,
• Shot Type,
• Aspect Ratio,
• Focus: normalized approximate coordinate (X, Y),
• Visual Intent,
• Key References,
• Product Angle,
• Background/Props,
• Lighting,
• Failure Avoidance.

Focus guidance:

• Hero images should not all have the same focus point.
• Detail images can follow a Z/F reading path for scroll behavior.
• White-background KV should keep product centered and clean.

Aspect Ratios

Default:

• hero/main images: 1:1 unless platform says otherwise.
• detail-page images: 3:4 for mobile scroll, 16:9 for banner/detail modules, or platform-specific.
• white-background KV: 1:1 with seamless white background.

If the current image tool cannot produce the exact ratio, disclose and use closest supported ratio.

Prompt Rules

Use English prompts by default for ecommerce image generation unless the this turn's tool list performs better with Chinese. Keep prompts concise, usually 120-200 words.

Start image-to-image prompts with the product reference:

<<<image_1>>> inherit the exact product form, proportions, surface finish, color, logo placement, and core structure.

Then describe:

1. camera distance and angle,
2. product placement and pose,
3. environment/props,
4. light direction and shadow behavior,
5. material detail,
6. ratio and quality,
7. negative constraints.

Default negative tail:

no product morphing, no distorted proportions, no messy reflections, no floating artifacts, no random on-screen text, no watermarks, no brand identifiers, no subtitles, no oversaturated colors, no physically impossible shadows, no ugly digital noise

For white-background KV add:

white seamless background only, no environment props, no colored cast

For feature/callout images add:

no decorative clutter outside designated callout zones

Do not include new marketing claims in generated text unless the user explicitly provided exact approved wording and the current image tool can render it reliably.

Visual Style Rules

Maintain:

• product contour consistency,
• product color consistency,
• material finish consistency,
• logo/text consistency,
• physically plausible lighting,
• restrained color palette,
• product category-appropriate props,
• platform readability.

Use a 90:10 color discipline:

• about 90% of the image should remain in the core visual palette,
• contrast/accent colors should support only the product or key selling point.

Avoid:

• random extra logos,
• visual claims not provided by user,
• oversaturated AI look,
• physically impossible shadows/reflections,
• props that contradict product category,
• product mirrored/flipped without reason.

Consistency Audit

After generation, review every image against the product baseline.

Audit dimensions:

• outline_proportion: product length/width and major silhouette.
• color_arc: dominant hue/saturation/lightness consistency.
• material_finish: matte/gloss/metal/glass/fabric behavior.
• logo_text: position, clarity, no invented marks.
• orientation_structure: facing direction and functional zones.
• background_pollution: white-background cleanliness and prop relevance.

Mark warnings if:

• product proportion visibly changes,
• color shifts noticeably,
• material finish changes,
• logo moves or mutates,
• product flips unexpectedly,
• white background has colored cast or props,
• scene props conflict with product tone.

If a warning affects the hero image or white-background KV, recommend regeneration before considering the kit complete.

Output Contract

Return:

• product_baseline: product visual and structural summary.
• ecommerce_spec: platform, ratios, style, color, claims, constraints.
• image_plan: hero/detail shot table.
• prompts: prompt per image.
• media: generated image URLs or pending states.
• audit: consistency report with pass/warning/fail.
• export_notes: ordered image kit usage.

Do not route to video unless the user explicitly requests video.