Eyeroll — AI skill for Claude Code

eyeroll

[](https://github.com/mnvsk97/eyeroll/actions/workflows/ci.yml) [](https://pypi.org/project/eyeroll/) [](https://pypi.org/project/eyeroll/) [](https://opensource.org/licenses/MIT)

AI eyes that roll through video footage — watch, understand, act.

eyeroll is a Claude Code plugin that analyzes screen recordings, Loom videos, YouTube links, and screenshots, then helps coding agents fix bugs, build features, and create skills.

Install

# Add the plugin to Claude Code
/plugin marketplace add mnvsk97/eyeroll
/plugin install eyeroll@mnvsk97-eyeroll

# Install the CLI
pip install eyeroll[gemini]      # Gemini Flash API (recommended)
pip install eyeroll[twelvelabs]  # TwelveLabs direct video understanding
pip install eyeroll[openai]      # OpenAI GPT-4o + OpenRouter/Groq/Grok/Cerebras
pip install eyeroll              # Ollama only (local, no API key) — requires Pillow
pip install eyeroll[all]         # everything

Setup

/eyeroll:init

Picks your backend, configures API key, and generates codebase context — all in one step.

For TwelveLabs directly:

export TWELVE_LABS_API_KEY=your-key-here
eyeroll watch ./bug.mp4 --backend twelvelabs

Commands

Command	What it does
/eyeroll:init	Set up eyeroll — pick backend, configure API key, generate .eyeroll/context.md
/eyeroll:watch	Analyze a video and present a structured summary
/eyeroll:fix	Watch a bug video → diagnose → fix the code → raise a PR
/eyeroll:history	List past video analyses

Usage

In Claude Code

You: /eyeroll:watch https://loom.com/share/abc123
     → Analyzes video, presents: what's shown, the bug, key evidence, suggested fix

You: /