sort-unsorted

To reclassify unsorted items in a markdown index, parse INDEX/sources.md and INDEX/tools.md, fetch external page metadata for enrichment, and re-apply taxonomy rules to update category rows. This process narrows classifications and generates a PR summary for review.

Automated tagging leaves ambiguous repository entries in an unsorted backlog because conservative classification rules avoid widening categories. A second-pass lookup fetches external repo metadata like descriptions and topics to enrich data and improve accuracy before reclassification.

The best way to clean up untagged GitHub repos is using a second-pass enrichment lookup that fetches repository topics and README excerpts. It re-applies category vocabulary heuristics to narrow classifications and opens a PR with uncertain cases marked for human review.

Yes, you can target unsorted sources or tools by specifying the target parameter. Running with default target=both processes up to max_rows=50 active rows where category=unsorted or kind=unsorted across both INDEX files.

No, the reclassification process never moves a fine category back to unsorted. It strictly narrows classifications by re-applying the standard taxonomy rules while keeping uncertain second-pass cases marked so humans can fix them.

A limitation of using repository metadata for topic enrichment is that ambiguous items may still lack sufficient details. The system handles this by keeping uncertain second-pass cases marked in the PR summary for human review rather than forcing a classification.