Webpage A4 Screenshot

Use when the user asks to screenshot a webpage, article, document page, Zhihu article, product page, or web content as images.

Default standard for 史老师:

• Capture only the main/body content area, not unrelated left/right sidebars, navigation, floating buttons, comments, or ads.
• Output clear large images, approximately A4 portrait.
• Prefer width 2480px output using CSS width 1240 and device scale factor 2.
• Do not cut text lines or images in the middle when avoidable.
• If an A4-height cut would split a paragraph, heading, list, table, or image, move the cut up/down to the nearest content-block boundary.
• Adjacent screenshots must have zero repeated content: next image starts exactly where previous image ended.
• Variable page heights are acceptable when needed to preserve content integrity.
• Verify zero overlap before sending.

Workflow

1. Use the helper script:

node skills/webpage-a4-screenshot/scripts/capture-a4-content.js '<URL>' /tmp/webpage-a4-capture

2. Check the summary:

• overlapCheck must be true.
• pages should be plausible for the article length.
• Output files are listed in files.

3. Send images with message as attachments.

4. If result is wrong:

• Too much chrome/sidebar: add or adjust content selectors in the script.
• Text/image still split: increase/decrease targetH, minH, or refine selector boundaries.
• Duplicate content: do not add bottom padding to clips; keep start = previous cut exactly.

Current helper behavior

The script:

• Opens Chromium headless through CDP.
• Loads and scrolls the page to trigger lazy content.
• Extracts a likely main article/content node.
• Re-renders content into a clean white article page.
• Computes cut points using block bottoms.
• Captures variable-height A4-like PNGs.
• Writes manifest.json with cut ranges and an overlap check.

Messaging note

For Weixin, send local files using absolute paths. Do not use relative paths.