screenshot-site

name	screenshot-site
description	Capture Playwright screenshots of a website under development. Use when the user asks to screenshot the site, grab screenshots of the app/UI, snapshot pages for visual review, or track visual evolution of a site.

Capture screenshots of the website in the current repo using Playwright and organise them into a dated archive plus a "latest" pointer folder.

Inputs

Accept these from the user (all optional):

• URL(s) — one or more URLs or paths to capture. If the user gives a path like /about, join it with the base URL.
• Base URL — defaults to http://localhost:3000. If that isn't responding, probe common dev ports (3000, 5173, 8080, 4321, 5000, 8000) and use the first one that returns 200. If nothing is running, ask the user for the URL or whether they want you to start the dev server.
• Viewport — defaults to 1920x1080. Accept shorthand like mobile (390x844), tablet (768x1024), desktop (1920x1080).
• Full page — default true. User can request viewport-only.
• Name — optional label for the screenshot file (e.g. homepage, dashboard). Default to a slug derived from the URL path (/ → home, /about/team → about-team).

Output layout

Write into ./screenshots/ relative to the current repo root (create if missing, and add a line to .gitignore if the user wants them ignored — ask once the first time you create the folder).

Structure:

screenshots/
├── MMDD/              # e.g. 0419 — dated archive for evolution tracking
│   ├── home.png
│   └── dashboard.png
└── latest/            # always overwritten with the most recent run
    ├── home.png
    └── dashboard.png

• MMDD uses the current local date, zero-padded (April 19 → 0419).
• Always write each screenshot both to the dated folder and to latest/ (overwriting).
• If a file with the same name already exists in the dated folder from an earlier run the same day, append a time suffix: home-1430.png.

Procedure

1. Resolve base URL. If the user gave one, use it. Otherwise probe localhost ports (above) with a quick curl -s -o /dev/null -w "%{http_code}" http://localhost:<port> and use the first 2xx/3xx response. If nothing's up, ask.
2. Resolve pages. If the user gave specific paths, use them. Otherwise default to just / (homepage). If the repo appears to be a multi-page site and the user said "screenshot the site" without specifying, ask which pages.
3. Prepare folders. mkdir -p screenshots/MMDD screenshots/latest.
4. Run Playwright. Prefer the Playwright MCP tools if available (mcp__plugin_playwright_playwright__browser_navigate, browser_resize, browser_take_screenshot). Otherwise write a short Node script using @playwright/test or playwright and run it with npx playwright — install on demand if missing.
5. For each page: navigate, set viewport, wait for network idle, take full-page screenshot, save to both screenshots/MMDD/<name>.png and screenshots/latest/<name>.png.
6. Report. List what was captured with file paths so the user can open them.

Node script template (fallback when MCP Playwright is unavailable)

const { chromium } = require('playwright');
const fs = require('fs');
const path = require('path');

const BASE = process.env.BASE_URL || 'http://localhost:3000';
const PAGES = (process.env.PAGES || '/').split(',');
const VIEWPORT = (process.env.VIEWPORT || '1920x1080').split('x').map(Number);

const mmdd = (() => {
  const d = new Date();
  return String(d.getMonth() + 1).padStart(2, '0') + String(d.getDate()).padStart(2, '0');
})();

const dated = path.join('screenshots', mmdd);
const latest = path.join('screenshots', );
fs.(dated, { :  });
fs.(latest, { :  });

( () => {
   browser =  chromium.();
   context =  browser.({ : { : [], : [] } });
   page =  context.();

   ( p  ) {
     url =  (p, ).();
     name = (p.(, ) || ).(, ) + ;
     page.(url, { :  });
     datedPath = path.(dated, name);
     latestPath = path.(latest, name);
     page.({ : datedPath, :  });
    fs.(datedPath, latestPath);
    .();
  }

   browser.();
})();

Notes

• Never commit screenshots automatically; leave that to the user.
• If the dev server isn't running and the user wants one started, check package.json for dev/start scripts and run in background — but ask first.
• For sites behind auth, ask the user how to authenticate (storage state file, login flow) before capturing protected pages.