MDBX Task Breakdown for Low-Capability Models

Version: MDBX-1-DRAFT

This document provides low-ambiguity, independently verifiable tasks.

1. Usage

Each assigned task should include:

• task ID
• required spec files
• non-negotiable constraints
• acceptance criteria

Recommended always-read files:

• 02.MDBX-Agent-Execution-Rules.md
• 01.MDBX-Product-Spec.md
• 03.MDBX-Storage-and-Sync-Spec.md
• 04.MDBX-Security-Spec.md
• 06.MDBX-SQLite-Schema-v1.md

2. Task principles

• one task, one primary objective
• every task must be testable
• no hidden context dependencies
• data model first, then IO, then migration, then UI

3. Core task list (summary)

1. establish SQLite v1 schema
2. implement vault initialization
3. implement project CRUD
4. implement entry CRUD
5. implement commit + parent relations
6. implement device head + branch head
7. implement attachment metadata MVP
8. implement small attachment inline mode
9. scaffold chunked attachment mode
10. implement tombstones
11. implement snapshot MVP
12. implement basic conflict detection
13. scaffold KDBX importer
14. scaffold KDBX exporter
15. implement Tiga mode switching (global/project/entry)
16. implement multiple unlock factors (PIN/password/security key)

Each task must define goal, constraints, forbidden changes, and acceptance checks.