Cleanup

The CLI still contains a legacy cleanup helper, but claudux update does not run an automatic semantic cleanup pass.

What exists today

lib/cleanup.sh provides:

• cleanup_docs — an interactive Claude prompt path that asks the model to look for obsolete docs (model-judged, not a deterministic scorer)
• cleanup_docs_silent — a no-op used during the main update path (explicitly empty so update does not silently delete files)

There is no built-in confidence threshold, no automatic file deletion pipeline with dry-run scoring, and no guarantee that obsolete pages are removed on every claudux update.

What to use instead

Prefer the write-boundary path Claudux actually ships:

1. Keep docs structure under the docs structure manifest
2. Prefer section-level updates with hash guards / source-boundary checks
3. Delete obsolete pages yourself (or ask the model via -m) and re-run npm test / claudux check as needed

Example:

claudux update -m "List obsolete pages that reference deleted APIs; do not delete yet"
# review, then delete intentionally

Why this page stays

The route remains /features/smart-cleanup so existing links keep working. The visible label is now simply Cleanup.

Related

• Content Protection
• Generation Pipeline
• Deterministic Generation