A Kubernetes LFX Mentorship project explored human-centered automation for localization maintenance in the AI era. Rather than automating translation, the focus was on building a Markdown-aware triage script for the kubernetes/website repository that detects drift between English source pages and localized versions. The script categorizes findings into triage levels (Orphan, Strong signal, Moderate signal, No signal) to help maintainers prioritize review work. The proposed workflow keeps humans in control: deterministic triage reports surface where attention is needed, optional AI assistance can help summarize context, but final decisions remain with localization reviewers. The core principle is that automation should protect reviewer attention, not consume it.
