I love my Xteink, it's such an intentional way of consuming content. I have a similar setup to OP. Here's my codex automation: > Find three diverse, substantiv

I love my Xteink, it's such an intentional way of consuming content. I have a similar setup to OP. Here's my codex automation: > Find three diverse, substantive Hacker News articles submitted within the previous 24 hours. Prefer readable linked articles rather than discussion-only, video, paywalled, or repository pages. For each selection, extract a clean UTF-8 XHTML-compatible article fragment and use dotepub's API to create an EPUB under public/books/ with a descriptive `Title - Author.epub` filename. Fix any EPUB 2 packaging or markup problems, including an invalid `xml:lang` on the OPF package element, and require every file to pass `epubcheck` with no errors before upload. Avoid duplicates already present locally or in the production OPDS catalog. Upload only the three newly created EPUB files to `hetzner:/opt/opds/public/books/`; never sync or delete the whole directory. Then verify production `/up` returns 200, unauthenticated `/opds.xml` returns 401, authenticated `/opds.xml` returns 200 and includes all three titles, each download returns `Content-Type: application/epub+zip`, and deployed copies pass `epubcheck`. Report the selected article titles and URLs, filenames, and verification results. Follow the repository AGENTS.md instructions and keep production credentials secret.