anf-website-builder

ETMA Plugins for Claude Code

A collection of open-source plugins for Claude Code — Anthropic's official CLI for AI-assisted software engineering.

Plugins

anf-website-builder

Build professional websites that don't look AI-generated.

Uses the ANIF Framework (Assemble, Normalize, Illustrate, Fill) to build landing pages from real designer components, real Unsplash photography, and autoresearch-validated design patterns. Developed through 12 iterations of automated scoring.

Features:

• Real components — shadcn/ui, Magic UI, and Aceternity UI instead of hand-written Tailwind
• Real photography — Downloads Unsplash photos as atmospheric section backgrounds (free, no API key)
• 30+ free image sources — Illustrations, 3D assets, animations, textures, gradients, device mockups
• Anti-AI-tell patterns — No default indigo, no Inter-only fonts, no round numbers, no generic headlines
• Per-page illustration — Automatically determines what visuals each section needs and sources them
• Content quality rules — Specific numbers, benefit-driven headlines, realistic testimonials
• Works with brainstorming — Pairs well with the superpowers:brainstorming skill for requirements gathering

Quick start:

claude plugin marketplace add ErikTMA/claude-plugins
claude plugin install anf-website-builder@etma-plugins

Then ask Claude to build a website:

Build me a landing page for my SaaS product
Create a website for a coffee subscription service
Scrape example.com and rebuild it with modern design

---

learn-from-video

Extract knowledge from any video — YouTube, Vimeo, Loom, or local files.

Give Claude Code a video URL and it "watches" it for you: extracts the transcript, detects scene changes to capture keyframes, and analyzes both text and visuals to produce structured knowledge you can recall later.

Features:

• Any video source — YouTube, Vimeo, Loom, or any site yt-dlp supports, plus local .mp4/.mkv/.webm files
• 3-tier transcript fallback — YouTube captions API → yt-dlp auto-subtitles → local Whisper AI transcription
• Scene-change keyframe extraction — detects visual changes using PySceneDetect, extracts key frames, deduplicates with perceptual hashing
• Video type classification — samples frames across the video, classifies content type (terminal, slides, talking head, screen recording), adjusts detection sensitivity per type
• GPU-accelerated Whisper — auto-detects NVIDIA GPU and VRAM, selects the best Whisper model for your hardware
• Docker-based — zero host dependencies beyond Docker. Two images: light (~500 MB) for download and keyframes, full (~8 GB) for Whisper + CUDA
• Persistent knowledge — saves transcripts, keyframes, and structured knowledge docs to ~/.claude/learned/, stores summaries in vector memory for future recall

Quick start:

claude plugin marketplace add ErikTMA/claude-plugins
claude plugin install learn-from-video@etma-plugins

/learn-from-video https://www.youtube.com/watch?v=...

Requirements:

• Docker (required)
• NVIDIA GPU + NVIDIA Container Toolkit (optional, for GPU-accelerated Whisper)

Installation

As a marketplace

Register the marketplace and install individual plugins:

claude plugin marketplace add ErikTMA/claude-plugins
claude plugin install anf-website-builder@etma-plugins
claude plugin install learn-from-video@etma-plugins

Manual

Clone and copy plugin files directly:

git clone https://github.com/ErikTMA/claude-plugins.git
cp -r claude-plugins/plugins/anf-website-builder ~/.claude/plugins/anf-website-builder

Contributing

Contributions welcome. Each plugin lives in plugins/<name>/ and follows the Claude Code plugin structure:

plugins/<name>/
├── .claude-plugin/
│   └── plugin.json          # Plugin metadata
├── skills/
│   └── <name>/
│       └── SKILL.md         # Skill instructions
├── references/              # Supporting documentation
├── commands/                # Slash commands (optional)
├── scripts/                 # Supporting scripts (optional)
└── README.md

License

MIT