larkin0302/group-daily-newspaper

This project is a Claude Code skill that turns a WeChat group chat from one day, or any chosen time window, into a printable newspaper laid out like the People's Daily, the Wall Street Journal, or other broadsheet papers. It is the printable companion to another skill called group-daily, which already makes a long magazine style image from the same chat history. The pitch is for people who want to print the result on real paper, frame it as a birthday gift or anniversary keepsake, or pin it to a wall as a piece of group culture. The output is strict about format. Each page is exactly A3 size at 96 dots per inch, which works out to 1123 by 1587 pixels. The number of pages is decided by how much actually happened that day, in even counts of 2, 4, 6, or 8 pages. Each page gets its own layout rather than a repeated template. Headlines and subheads are rendered with a Chinese friendly word break setting so lines never end on a half word, and font sizes are picked dynamically between 22 and 42 pixels based on the character count. Installation is two git clones into the Claude skills folder. The newspaper skill needs the older group-daily skill already installed, because it consumes the story.json and avatars.json files that group-daily produces. The other requirements are Chrome or Chromium for the headless PDF print step and Python 3.10 or newer. A typical run copies a blank layout plan template, lets the AI fill it in with page templates for the day, runs a Python script to render HTML, forces a fixed page height with a sed command to stop Chrome from splitting pages, prints to PDF with Chrome headless, then verifies the output with pdfinfo to confirm A3 sizing. Four page templates are provided. The masthead template is always page one, with a hero story, briefings, a photo strip, and day statistics. The communal template covers shared output. The feature template handles the magazine style supplement. The cast template is a personality highlight reel, recommended for the last page. Everything runs locally. The skill ships only blank examples, with a gitignore that blocks real chat data, and the author suggests keeping live data in a separate vault folder outside the repo. The skill is MIT licensed.