Agent Skill Specification

Skill

A model-agnostic Skill that lets any AI assistant take screenshots and record videos of any public URL. Works with any tool that supports the open Agent Skills spec — Claude Code, Cursor, Cline, Codex CLI, Gemini CLI, Roo Code, Zed — and falls back to a wrapper file for tools that don't.

Install the Skill

Run the 1-line clone command for your CLI or code IDE:

bash — Claude Code

git clone https://github.com/Website-Screenshot-Online/websitescreenshot-skill.git ~/.claude/skills/websitescreenshot-skill

Installs globally across all your local projects and terminal sessions.

Try asking your AI assistant:

“Take a full-page screenshot of https://news.ycombinator.com in PNG format.”

“Record a 5-second scrolling MP4 of https://github.com and check video status.”

“Capture a mobile screenshot of https://stripe.com with ads blocked.”

Configure API Key

Export your API key so your assistant can call the API:

export SCREENSHOT_API_KEY=ws_your-api-key

Get API Key

Add this line to your ~/.zshrc or ~/.bashrc to persist across terminal sessions.

Tools Available to Your AI

Three tools are exposed with the same capabilities as the REST API:

take_screenshot

Capture a PNG/JPEG/PDF screenshot of any public URL.

PNGJPEGPDF

record_website_video

Record a scrolling webm/mp4/gif of any public URL (sync or async).

MP4WEBMGIF

check_video_status

Poll the status of an async video job. Results are kept for 24 hours.

Async Polling24h TTL