muapi-ad-creative
Generate a high-converting ad creative set — hero image, ad copy variations, and platform-optimized crops for Meta, Google Display, and LinkedIn.
适合你,如果你需要快速制作高转化率的广告图片和文案
用别的 agent?下载 .zip 解压,把文件夹放进它的技能目录
~/.claude/skills/(项目级 .claude/skills/)~/.codex/skills/npx oh-my-skill add samuraigpt/generative-media-skills/muapi-ad-creativecurl -fsSL https://oh-my-skill.com/install.sh | bash -s -- samuraigpt/generative-media-skills/muapi-ad-creativenpx oh-my-skill verify samuraigpt/generative-media-skills/muapi-ad-creative怎么用
商店整理自技能原文 · 版本 a1c4c98 · 表述以原文为准生成广告创意素材:先创建一张主视觉图和三条文案变体,等你选定文案后,再自动裁剪出适配 Meta、Google Display、LinkedIn 的多种尺寸图片。
当你提到“广告创意”、“Facebook 广告”、“Google 广告”等关键词,或要求制作广告素材时触发。
技能原文 SKILL.md
Ad Creative Set
Generate a high-converting ad creative set — hero image, ad copy variations, and platform-optimized crops for Meta, Google Display, and LinkedIn.
Inputs
| Name | Type | Required | Default | Description | |:---|:---|:---|:---|:---| | product_or_service | text | yes | — | What is being advertised (e.g. "SaaS project management tool for remote teams"). | | target_audience | text | yes | — | Who the ad is for (e.g. "startup founders aged 25–40, tech-savvy"). | | campaign_goal | text | no | awareness | Campaign objective — "awareness", "consideration", or "conversion". | | tone | text | no | professional, clean, modern | Creative tone and visual style (e.g. "bold and disruptive", "luxury minimal", "friendly and approachable"). | | product_image | image_url | no | — | Optional product or brand image URL already in the session. |
Steps
This skill has TWO phases. Phase A creates the hero concept for approval; Phase B fans out to platform formats.
Phase A — Hero image + Ad copy
Submit ONE the plan with:
- Hero image —
muapi image generate(model=nano-banana-pro) ormuapi image edit(model=nano-banana-pro-edit) if{{product_image}}is provided: - Aspect ratio: 1:1 (universal starting point).
- Prompt must capture: product/service benefit, target audience lifestyle cue, campaign tone.
- Style:
{{tone}}, advertising photography, clean background, product focus, ultra detailed, commercial quality. - Tier: quality.
After the plan executes, present the hero asset and 3 ad copy variations:
- Variation A — Problem-aware hook: "Tired of X? [Product] fixes that."
- Variation B — Benefit-led: "[Feature] → [Outcome] for [Audience]."
- Variation C — Social proof / urgency: "X teams already use [Product]."
Each variation includes: Headline (6 words max), Body (20–30 words), CTA button text.
Ask which copy variation to use for Phase B. Wait for user confirmation.
Phase B — Platform crops
Once the user picks a copy direction, submit a SECOND the plan with parallel crops:
muapi image edit→ 1:1 (Facebook/Instagram feed, 1080×1080)muapi image edit→ 9:16 (Story/Reels, 1080×1920)muapi image edit→ 1.91:1 (Facebook feed wide, 1200×628)muapi image edit→ 1:1 (LinkedIn feed, same as FB)
For each crop:
- Prompt: "Reframe for [platform] ad format. Keep product/subject centered and uncropped. Maintain original palette and tone. Leave headroom/footroom for text overlays."
- All crops run in parallel.
Return one asset per format with the recommended copy overlay placement for each.
Notes
- If
campaign_goalis "conversion", emphasize urgency and direct CTA in copy. - If
campaign_goalis "awareness", prioritize visual impact over text density. - Reference
product_imagevia$nX.urlsyntax in Phase B nodes to ensure consistency. - Do NOT auto-confirm Phase B without user picking a copy variation.
Trigger Keywords
ad creative, advertisement, facebook ad, meta ad, google ad, linkedin ad, paid ad, ad banner, display ad
Notes for the Executing Agent
- This recipe is LLM-orchestrated: read each phase, gather any missing inputs from the user, then call
muapiCLI commands. Usemuapi auth configurefirst ifMUAPI_API_KEYis unset. - For model IDs without a CLI alias yet, fall back to the raw endpoint via
curl -X POST https://api.muapi.ai/api/v1/<endpoint> -H "x-api-key: $MUAPI_API_KEY" -H 'content-type: application/json' -d '{...}'and poll withmuapi predict wait <request_id>. - Substitute
{{input_name}}placeholders with the user's actual inputs before issuing each call.