用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt;运行前请先检查来源。
npx skills add https://github.com/LYL1015/JarvisHub --skill flova-product-commercial-short命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地?可下载 SkillsMP 当前能够提供的文件。
正在显示 SKILL.md
Use when the user asks JarvisHub AI Chat or a `webHero` workflow to generate a website, landing page, homepage, brand-style web design, preview-first webpage workflow, or wants section screenshots, webpage assets, and final HTML to be produced as one staged flow.
把 assets/demo 提炼成运行时可用的视觉连续性方法论:先锁资产/角色/场景/镜头语义,再做扩镜与视频。(素材生成场景:角色一致性、视觉锚点锁定、单素材生成前的锚点规范与一次只改一个变量。)
基于用户提供的设计稿图片还原真实网页。该 skill 可用于任意前端项目,但强依赖本地已启动且已授权的 JarvisHub vision 工作流;视觉模型由服务端默认配置决定,支持外部传入 vision prompt,且不可静默降级。
基于 SOC 职业分类
| name | flova-product-commercial-short |
| description | Use when 用户基于上传的产品图片制作商品宣传短片、电商产品视频、产品在使用场景中的商业广告、功能氛围混剪或需要产品元素、场景元素、镜头视频逐步确认的短片。 |
用于商品宣传短片:从用户上传的产品图出发,设计产品展示、使用场景、功能氛围镜头与商业节奏,逐步生成元素资产、视频片段和最终合成计划。
This workflow is a general product commercial workflow. It is lighter than flova-visual-tvc-campaign and more narrative/environmental than flova-beat-cut-motion-video.
imageUrl / videoUrl,不能把提交态当完成态。critic sub-agent:只读取真实媒体并评审,不生成、不补素材。blocked 项,除非本轮工具列表明确暴露对应能力。Use this skill when the user wants:
Use a more specific skill when applicable:
flova-jewelry-12-shot-adflova-fashion-showcase-shortflova-industrial-product-promoflova-beat-cut-motion-videoflova-visual-tvc-campaignFollow the JarvisHub Execution Model for asset binding, image generation, video generation, review, and assembly. Audio is a plan or bound user asset unless this turn exposes a real audio generation tool.
If current tools can generate scene images, video clips, or assemble timelines, use them through the proper sub-agent. If audio generation is not exposed, return an audio plan instead of claiming generated audio.
Collect or infer:
product_reference: required. Use uploaded product image if available.product_category: SKU type and key selling points.aspect_ratio: default 9:16 for social commerce, 16:9 for brand/website video, 1:1 or 4:5 for feed ads.duration: default 15-30s unless user specifies.style: premium, energetic, lifestyle, technical, outdoors, minimalist, etc.platform: ecommerce detail page, Douyin/TikTok, Xiaohongshu/Reels, website hero, paid social.must_show: logo/package, feature, material, usage scene, price/promo text if allowed.forbidden: unauthorized logo/text, claims, unsafe usage, style taboos.audio: BGM, VO, SFX, or silent.If the product image is missing, ask for it before media generation. You may still draft a generic plan, but do not claim product-faithful output.
Pause for confirmation after important stages in interactive runs:
For autonomous agent runs, replace user confirmation with critic review and make the decision explicit.
A good product commercial should not be only "a package on a table." Mix two shot types.
The product appears clearly in a real or stylized usage scene.
Use for:
Examples:
The product may be absent or partially visible, but the shot sells the benefit.
Use for:
Examples:
The storyboard should alternate shot scale and purpose:
For every shot include:
shot_idtype: product-in-environment or support/feature/atmospherescene_elementproduct_presence: full, partial, macro, absentcommercial_pointframingcamera_movementactionlighting_materialaudio_or_sfxtransitionduration_targetPrefer short, purposeful shots for energy. A shot can be 5-15s if it contains internal beats, but each internal beat must have a clear visual change.
Use varied camera grammar:
Keep the product visually faithful: structure, color, material, package markings, and recognizability should not drift.
Before generation:
Do not regenerate the product itself if the uploaded image is already usable. Generate product views or clean asset boards only when required by the video model or user request.
Useful element asset categories:
Product_RootScene_UseCase_01Scene_UseCase_02Macro_Material_DetailRecurring_PropEnd_Card_BackgroundAvoid per-shot random assets; they break continuity and product identity.
Use Chinese prompts when the user is working in Chinese, while retaining necessary model tokens and negative phrases if useful.
Each prompt should follow:
For product-in-environment shots:
以产品参考图作为唯一产品外观依据,保持产品结构、颜色、材质和包装识别点一致。镜头展示产品在[使用场景]中的真实部署,[摄像机运动],[动作/功能表现],[光影材质],突出[商业卖点]。无字幕、无随机文字、无水印、no music。
For support/feature/atmosphere shots:
围绕产品卖点[功能/情绪]设计氛围镜头,[场景动作],[镜头运动],[材质/光影],画面服务于产品价值暗示。若产品不出现,明确保持该镜头为辅助功能/情绪镜头。无字幕、无随机文字、无水印、no music。
If the shot has separate VO or narration, do not put narration text inside the video prompt unless the current model is expected to generate speech.
Plan at least one global BGM direction unless the user asks for silent output. Bind a real BGM asset when the user provides one; generate one only when this turn exposes an audio tool.
For VO:
For product ads, useful SFX include:
Do not let generated video clips create uncontrolled music if audio will be assembled separately.
Assemble in the storyboard order.
Maintain:
Respect brand or platform requirements:
If timeline assembly is unavailable, return ordered clip URLs and an assembly plan with shot durations and audio placement.
Check:
If a clip fails product fidelity, narrow the prompt around the product reference and regenerate before changing the storyboard.
Return:
spec: title, ratio, duration, style, platform, language.product_analysis: product category, selling points, visual constraints.storyboard: shot table with shot type and commercial function.assets: uploaded/reused/generated assets and status.prompts: image prompts and video prompts.media: generated URLs or pending states.audio: BGM/VO/SFX plan, bound user assets, or generated assets only when a real audio tool is exposed.assembly: final video URL or ordered assembly plan.review: issues, regeneration needs, and what was verified.Do not treat a prompt table or storyboard as a generated commercial.