Cross-Ref Topic Ledger

Workflow

Per candidate:
- [ ] Step 1: Parse topic-ledger.md into {slug, title, status, tags, last_touched} records
- [ ] Step 2: Extract 2-5 topic tags from the candidate's summary (reuse ledger vocabulary)
- [ ] Step 3: Match candidate to ledger entries by tag overlap (≥2 shared tags) OR semantic title similarity
- [ ] Step 4: Classify: NEW | OVERLAPS seed:{slug} | OVERLAPS draft:{slug} | OVERLAPS published:{slug}
- [ ] Step 5: If OVERLAPS published, generate reinforcement_angle (1 line)
- [ ] Step 6: Annotate candidate with dedup_status, overlap_slug, reinforcement_angle

Classification rules