fix: pin SEO URLs to code-echo.pages.dev on both hosts

[assertion]

Summary

GitHub Pages was shipping relative canonical/hreflang/feed links, relative sitemap <loc> values (invalid), and a robots.txt with no Sitemap line. Cloudflare production was already fine because SITE_URL is injected at build time. This makes both builds emit absolute URLs pointing at the canonical host https://code-echo.pages.dev, so the GitHub Pages copy consolidates ranking signals to Cloudflare.

What changed

• Set url: https://code-echo.pages.dev in _config.yml (local serve still overrides via _config_local.yml). Cloudflare preview should leave this as-is so preview hostnames are not indexed as a second origin; relative CSS/nav links keep previews browsable.
• Shared _includes/absolute-origin.html fallback: if site.url is empty or is assertion.github.io, templates still emit the Cloudflare origin.
• Custom sitemap.xml (Liquid) so GitHub Pages — which cannot run _plugins/ — still writes absolute <loc> values, keeps lastmod on posts, and omits empty paginator pages / 404 / the markdown test page.
• robots.txt always includes Sitemap: https://code-echo.pages.dev/sitemap.xml on production builds.
• Optional google_site_verification / bing_site_verification in _config.yml (empty). Rendered as google-site-verification and msvalidate.01 only when set.
• hreflang="x-default" now always points at the ZH twin.
• /404.html and /markdown are noindex + sitemap: false.

_posts/ untouched.

Verification (Jekyll 3.10 local builds)

Production (jekyll build --config _config.yml) — same committed url used by GitHub Pages and Cloudflare:

User-agent: *
Allow: /

Sitemap: https://code-echo.pages.dev/sitemap.xml

• sitemap.xml: 63 absolute https://code-echo.pages.dev/... locs, 0 relative, 57 lastmods (all posts). Excluded /page4/, /markdown, /404.html. Kept /, /about, /categories/, /wiki/, /page2/, /page3/.
• ZH post: <link rel="canonical" href="https://code-echo.pages.dev/2026/10/09/ai-twitter-hots" /> plus matching hreflang / og:url / feed / .md / JSON-LD.
• EN post: hreflang="x-default" points at the ZH URL.
• 404, /markdown, /page4/ each have robots=noindex.
• Empty verification keys emit no meta tags. Overlay google_site_verification / bing_site_verification rendered google-site-verification and msvalidate.01.

Build	canonical (2026-10-09 ZH)	robots Sitemap
_config.yml (prod / GH / CF)	https://code-echo.pages.dev/2026/10/09/ai-twitter-hots	https://code-echo.pages.dev/sitemap.xml
+ _config_local.yml	http://localhost:4000/2026/10/09/ai-twitter-hots	http://localhost:4000/sitemap.xml
+ empty url: "" fallback	https://code-echo.pages.dev/2026/10/09/ai-twitter-hots	https://code-echo.pages.dev/sitemap.xml

After merge

1. Paste Google / Bing verification codes into _config.yml when issued (google_site_verification, bing_site_verification).
2. In Google Search Console and Bing Webmaster Tools, add https://code-echo.pages.dev as the property and submit https://code-echo.pages.dev/sitemap.xml. site: will stay empty until the property is verified and crawled.
3. Cloudflare preview: do not overlay url with the preview hostname, or preview pages will advertise a second origin.

Noted, not changed

• GitHub Pages still cannot run _plugins/, so plaintext .md copies are Cloudflare-only. After this change the rel=alternate markdown links on the GitHub Pages HTML point at the Cloudflare URLs, which do exist.
• GitHub Pages https_enforced is currently off (http://assertion.github.io). Harmless once canonicals point at Cloudflare.

  

[cloudflare-workers-and-pages]

Deploying code-echo with    Cloudflare Pages

Latest commit:	2d31964
Status:	✅  Deploy successful!
Preview URL:	https://c7089680.code-echo.pages.dev
Branch Preview URL:	https://cursor-canonical-domain-seo.code-echo.pages.dev

View logs