Skip to content

Journal Backlog Processing Directions

How to run Running Notes backlog pipeline — cluster review, routing proposals, rewrite presentation

When to use

user wants to process or clean up their Running Notes backlog

Slash command: /wb-journal-backlog

Linked workflow: daily-journal/process-backlog

  • journal/running_notes
  • journal/vault_write_at_location

Directions

Start via mcp__work-buddy__wb_run("process-backlog"), then advance with wb_advance.

Default to today's journal unless a different date is specified.

Cluster review rules

  1. High confidence (cohesion >= 0.7): Accept unless obviously wrong
  2. Medium confidence (0.5-0.7): Check thread summaries -- do they belong together?
  3. Low confidence (< 0.5): Likely need splitting
  4. Uncertain edges with strength > 0.55: Seriously consider moving that thread
  5. Singletons: Should any join an existing cluster?

Present: "I found N clusters. M look correct, K need your input."

Routing proposals

For each unit: use semantic_search to find relevant vault location, use context for type guesses, flag cryptic items as low confidence, note staleness for completed work.

Rewrite presentation

Before rewriting Running Notes, present before/after summary. Thread ID tags are removed (processing artifacts). This is a filesystem write -- consent required.

Don'ts

  • Don't run in the background -- requires active participation
  • Don't skip cluster review -- bad grouping cascades into bad routing
  • Don't auto-route without user confirmation
  • Don't pressure to process everything -- partial processing is fine
  • Don't restore carried-over banners
  • Don't create new organizational systems during this workflow
  • Insight hoarding: this workflow directly addresses it. Start with easy items to build momentum
  • Infrastructure displacement: if user redesigns task system mid-triage, name it and redirect
  • Branch explosion: routing may reveal many live threads. Don't solve that here -- make it visible