Consolidation No Action — 26 June 2026
What was reviewed
The raw directory under /opt/data/wiki/raw/ was scanned for near-duplicate clusters (threshold >= 85% similarity) from the last 30 days. The full file inventory:
| Directory | Files | Notes |
|---|---|---|
agent-tool-watch/ | 3 | Tool monitoring entries |
cluster-janitor-dryrun.md | 1 | One-off dry run report |
mercury-learning-check-*.md | 2 | Daily learning checks (identical empty templates — trivial, not worth merging) |
queue-baseline/ | 24 | Daily queue observability snapshots |
skill-bench-*.md | 3 | Weekly skill candidate surveys |
verifier-log.md | 1 | Paris verifier audit log (append-only, keep as-is) |
web-captures/ | 1 | Single queue observability capture |
Decision: No cluster meets consolidation threshold
No cluster of near-duplicates was found. The largest repetitive set is queue-baseline/ with 24 files, but these are daily cron outputs by design — each file captures a unique date’s snapshot. They share a template structure but carry distinct time-series data. Collapsing them would lose the daily granularity that the observability pipeline depends on.
The two mercury-learning-check files (2026-06-24, 2026-06-25) are identical empty templates, but merging a pair of trivially empty files doesn’t justify an MR or concept page. The skill-bench reports follow the same pattern — routine cron output with unique dates.
What was NOT consolidated
Per constraints: entities/ and email/ were excluded. No sessions or project session notes had near-duplicate clusters either.
Conclusion: No consolidation action needed this run. The raw directory is healthy — files are distinct or intentionally repetitive (daily cron). Next review recommended in 7 days.