What problem does it solve?
Provides a lightweight, convention-first project management system that preserves conversational context across AI-human sessions, prevents information loss from context compaction, and centralizes project state for discoverability and continuity.
Core Features & Use Cases
- Three-tier context model: Working memory (CONTEXT.md), semantic memory (REFERENCE.md), and episodic memory (sessions/) to keep active work concise and searchable.
- Single source of truth index: projects/index.yaml stores project metadata and status so folders remain stable regardless of lifecycle changes.
- Archival and session protocol: Line-count budgets, two-phase archival commit, and session start/end checklists that prevent lost details during compaction and multi-session handoffs.
- Cross-project lessons: _lessons/ with date-prefixed files for discoverable patterns and incident reports that agents can grep and reuse.
Quick Start
Create a new project by adding an entry to projects/index.yaml, creating the project folder with a CONTEXT.md summary, and run the session start checklist to capture initial state.