This skill extracts a self-contained HTML file from a web application. Its default strategy launches headless Chrome through a snapshot script, captures the rendered DOM and computed CSS rules, and inlines images and icon fonts as base64.
A second strategy uses a browser subagent for pages that need clicks or form input first. A last-resort fallback flattens React components into a mock page.
What You Get
- A single HTML file per route with styles and images inlined
- Snapshot script flags for wait time, viewport, html class and title
- Authentication script support for login-gated pages
- Framework notes with recommended wait times
Common Use Cases
- Capturing pages of a local app to upload to Stitch
- Sharing a static version of a page without running the app
- Saving a specific UI state such as dark mode or a chart view
What Makes It Unique
It resolves styles from the live stylesheet rules, so dynamic Tailwind and CSS-in-JS output survives, and it handles srcset images, icon fonts and canvas charts. Two checkpoints make the agent confirm the strategy and the running app before it launches a browser.
Under The Hood
The agent runs a bundled TypeScript snapshot script through npx tsx. It launches headless Chrome with Puppeteer, loads the local URL, optionally runs an auth script, and writes the HTML file to disk. A second script converts a flattened mock page, and a post-process script inlines local images. The agent starts and stops the local dev server.