What problem does it solve? Processing a cluttered Obsidian inbox with dozens of mixed items (web clippings, voice memos, VTT files, PDF attachments) is slow and error-prone when done one item at a time, and reading all content in a single session overflows the AI's context window. ## Core Features & Use Cases - Unified Inbox Processing: Handles all content types (clippings, transcriptions, VTT files, attachments) in one workflow, routing each item to the correct analyzer based on its type. - Parallel Subagent Architecture: Spawns isolated worker agents in batches of 10 to enrich, analyze, and create notes, keeping the coordinator context under 25k tokens even for 50+ items. - Crash-Resilient Resume: Subagents persist proposals immediately via task metadata, so an interrupted session can resume exactly where it left off. - Collaborative Review Checkpoint: Presents all proposals in a single table where you can accept, edit, or delete items before originals are cleaned up. - Use Case: You return from a week of saving articles, recording voice memos, and clipping tweets. Run the triage skill to process all 50 inbox items, review the proposed areas and projects in one table, and approve cleanup in a single interaction. ## Quick Start Ask the assistant to run the triage skill to process everything in your Obsidian inbox, optionally filtering by type such as clippings, voice memos, or a specific filename.
