跳到正文
cccskills
免费在 GitHub 公开

Product Video Ads

Produce product-led video ad concepts and variants while preserving approved product details, claims, and platform constraints.

中文简介正在准备中。当前显示原文说明。

实测评分 未实测

查看安装方法

包含的文件(1)

  • SKILL.md4.2 KB

SKILL.md(原文)

安装前可以查看交给 AI 代理的指令内容。

Product Video Ads

Mission

Turn an approved product brief into reviewable video-ad variants. Protect the product's geometry, packaging, label text, color, usage mechanics, and approved claims while leaving room for camera, set, pacing, and art-direction choices. This skill creates creative drafts; it does not make legal claims or launch a campaign.

Required inputs

  • product/SKU identity, hero features, approved claims, and prohibited claims;
  • authorized product images or videos, angles, dimensions, and packaging invariants;
  • audience, offer, CTA, language, platform, duration, ratio, and caption needs;
  • brand guide, visual tone, creator/presenter requirements, and music policy;
  • desired variants, budget, review owner, and definition of done; and
  • rights state for product, logo, talent, location, music, and supplied assets.

Never fill a missing benefit, price, testimonial, certification, or comparison with plausible marketing copy. Mark it as unapproved and ask for the source.

Phase 1: lock the product

Create a product preservation block:

product_id / SKU:
approved reference roles:
must preserve: silhouette, proportions, materials, colors, labels, logo,
                 controls, ports, included parts, and orientation:
approved claims and exact wording:
prohibited or unverified claims:
allowed transformations: set, lighting, camera, motion, background:
rights / review owner:

Use one or more clear references and record their roles. If packaging text or small label detail is essential, plan a composited text/packaging review rather than trusting an unverified generated render.

Phase 2: develop the hero

  1. Normalize the ad into hook, product reveal, proof/demo, benefit, and CTA beats. Keep the actual claim source beside each beat.
  2. If no adequate product reference exists, use the adjacent image agent or an available media.generate_image workflow to propose a still. Obtain review before animating it; do not silently replace a supplied product.
  3. Choose an I2V or reference-capable video schema from MODELS.md when product fidelity matters. Resolve its live fields through muapi-video-tools.md.
  4. Generate the hero shot first. Review silhouette, label, scale, hand/product interaction, reflections, and motion before making variants.

Keep the product preservation block separate from style instructions. Prompt for observable camera and action; do not rely on a model to infer the product from a brand name alone.

Phase 3: variants and delivery

Generate controlled variants by changing one meaningful factor at a time: hook, setting, opening camera move, presenter treatment, proof beat, or CTA layout. Keep the approved product parent fixed when making downstream edits. Use Video QA for product fidelity, claims, temporal stability, text, audio, caption, and framing checks.

Create platform crops or captions only after selecting the hero. State whether audio is native, supplied, separately voiced, or silent. If narration is needed, hand off to the voice agent or use an explicitly available speech capability; do not invent a voice authorization.

Output contract

Return a variant table with hook, product reference roles, model/endpoint, prediction ID, output URL, status, exact claim source, audio treatment, ratio, QA findings, and parent asset. Include unapproved claims, failed jobs, and remaining legal/brand review. Save manifests and receipts under .video/ only when permitted, following reports-and-receipts.md.

No output is a legal, performance, or platform-compliance guarantee. Do not publish or launch spend from this skill.

评价

还没有评价。欢迎分享你的使用感受。

评论

无需登录即可发表。欢迎分享使用感受、提问,或说明在哪些环境下无法运行。

还没有评论。来写第一条吧。

提交后会立即公开。请勿填写邮箱、电话等个人信息。URL 最多 2 个。

0 / 1000 字

此表单受 reCAPTCHA 保护,适用 Google 的隐私权政策和服务条款。

同一仓库的技能

Plan draft video-ad variations from supplied performance evidence, preserving the distinction between observed data and creative hypotheses.

中文简介正在准备中。当前显示原文说明。

SamurAIGPT/open-ai-video-agent82026年8月28日 更新

Create authorized presenter, avatar, and UGC-style video drafts with explicit voice, likeness, consent, and synthetic-media handling.

中文简介正在准备中。当前显示原文说明。

SamurAIGPT/open-ai-video-agent82026年8月28日 更新

Analyze a script or existing edit, identify visual gaps, and generate supplemental B-roll coverage with clear parent and factual-status labels.

中文简介正在准备中。当前显示原文说明。

SamurAIGPT/open-ai-video-agent82026年8月28日 更新

Build a consented continuity system for an authorized person, fictional character, product, or world across video shots.

中文简介正在准备中。当前显示原文说明。

SamurAIGPT/open-ai-video-agent82026年8月28日 更新

Translate creative intent into coherent cinematography, framing, camera, lens, lighting, motion, and sound direction for a video shot.

中文简介正在准备中。当前显示原文说明。

SamurAIGPT/open-ai-video-agent82026年8月28日 更新

Create a non-explicit romantic scene from authorized adult references with clear consent, identity boundaries, and synthetic-media labeling.

中文简介正在准备中。当前显示原文说明。

SamurAIGPT/open-ai-video-agent82026年8月28日 更新

查看 SamurAIGPT 的全部技能

报告此技能的问题