motionsites-promptcapture
“Don't start from a blank prompt. Let MotionSites show you the design language — then make it yours.”
A Codex skill that turns “I need a website with great UI design” into a curated MotionSites prompt and a concrete implementation plan.
Stuck on the visual direction for your landing page, SaaS site, or portfolio?
Copying generic AI prompts and getting generic results?
Wondering how to turn a beautiful animated-site prompt into something that fits your stack?
This skill goes to motionsites.ai, finds the prompt that actually matches your request, explains why it fits, and gives you a plan to adapt it to your project.
Install · Quick start · Workflow · Capabilities · Companion skills · Directory · Principles · Limitations · Compatibility · 中文版
Installation
Method 1: Codex skill-installer
python "${HOME}\.codex\skills\.system\skill-installer\scripts\install-skill-from-github.py" --repo bingo33-gif/motionsites-promptcapture --path . --name motionsites-promptcapture
Method 2: Clone into your skills directory
git clone https://github.com/bingo33-gif/motionsites-promptcapture.git "${HOME}\.codex\skills\motionsites-promptcapture"
Restart Codex (or open a new task) and the skill will be discovered automatically. More details in INSTALL.md.
Quick start
- Copy this repository link to Codex and let it install the skill for you.
- Open your project folder in Codex.
- Say: “I need you to create a website and use UI design.”
- Tell Codex the site type, visual style, and tech stack (or let it ask); the skill will find the matching prompt and present an optimization plan.
The skill only starts implementing after you confirm the plan.
Recommended companion skills
figma(plugin) — pull design context from Figma when you already have a design filefrontend-app-builder(Build Web Apps plugin) — implement the approved plan with high-quality UIplaywright/frontend-testing-debugging— verify animations, responsiveness, and console errors in a real browser
Current capabilities
- Requirement extraction: site type, visual style, animation needs, tech stack
- MotionSites prompt discovery: category + style + motion matching
- Verbatim prompt output with a matching rationale
- Optimization plan: stack migration, animation guidance (GSAP ScrollTrigger + Lenis), visual assets, browser verification
Recommended workflow
- Clarify the request.
- Browse motionsites.ai and select 1–3 matching prompts.
- Present the prompt(s) with a rationale.
- Deliver the optimization plan.
- Implement only after user confirmation.
- Verify the result with browser testing.
Directory structure
motionsites-promptcapture/
├── SKILL.md
├── INSTALL.md
├── README.md
├── README.zh-CN.md
├── LICENSE
├── agents/
│ └── openai.yaml
└── .github/
├── scripts/
│ └── quick_validate.py
└── workflows/
└── validate-skill.yml
Design principles
- Never fabricate a prompt — only show what actually exists on motionsites.ai.
- Keep the original prompt verbatim; optimize the plan, not the source text.
- Match the prompt to the user's real stack and constraints.
- Recommend browser verification before calling a UI done.
Current limitations
- motionsites.ai is a JavaScript-rendered site; browsing it may require the in-app browser.
- The skill reads the live site on each use rather than relying on cached content, so prompts can change as the site evolves.
- The skill delivers prompts + plans; it does not generate sites by itself.
Compatibility
- Codex: native support.
- Other Agent Skills-compatible clients (Claude Code, Cursor, etc.): the
SKILL.md+agents/structure is portable.
Contributing
A GitHub Actions workflow (.github/workflows/validate-skill.yml) validates SKILL.md and agents/openai.yaml on every push and pull request. Run the same check locally:
pip install pyyaml
python .github/scripts/quick_validate.py .
License
MIT
No comments yet
Be the first to share your take.