macrodata-distill

To extract actionable knowledge from daily conversations, this Skill processes your OpenCode sessions by spawning a sub-agent per conversation to prevent context blowup. It aggregates distilled actions, facts, and decisions into a structured JSON payload.

A sub-agent prevents context blowup by individually processing each conversation transcript, preserving fidelity. The coordinator then aggregates these isolated results into a single JSON payload, preventing the main context window from overflowing.

Yes, you can write distilled facts and decisions to an existing journal. The coordinator aggregates the extracted insights and writes updates directly to the macrodata journal and corresponding entity files.

Yes, this approach works for consolidating memory across many daily sessions. It distills ongoing conversations into compact, searchable knowledge by processing each session independently before aggregating the results.

The coordinator outputs a JSON payload when distilling conversations. This structured payload contains categorized arrays for distilled_actions, facts, and decisions extracted from the processed sessions.