link-rot

Build the site, then run a script that extracts every external href from the prerendered HTML and probes each one with a HEAD-then-GET fallback. Verify any reported failures manually in a real browser before treating them as dead, since many hosts block scripted requests.

Many hosts answer scripts with 403, 429, or 999 status codes while serving real browsers fine, so a probe failure is only a lead. Retry with curl and a browser user agent, then load the URL in an actual browser before calling a link dead.

A build can validate internal links and make broken internal hrefs impossible, but it offers no guarantee for external URLs. External rot requires a separate scheduled loop that probes outbound links and verifies failures by hand.

Only swap an href for a clear new home or remove the dead anchor with minimal wording changes. When the replacement requires a judgment call, such as choosing which article now makes the point, record it as a proposal for human approval instead of guessing.

Maintain a known-blockers list of hosts that reject scripted requests on principle, with a written reason for each entry. Spot-check those links manually on every run, since the script never probes them and the manual pass is their only coverage.