rose-memory

Store project-local checkpoints, requirements, and evidence-backed findings in a SQLite database.

2|Updated May 3, 2026
One-click install
npx skills add https://github.com/Rosetears520/aili-workflows --skill rose-memory
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: rose-memory
Source: https://github.com/Rosetears520/aili-workflows/tree/main/skills/rose-memory
Command: npx skills add https://github.com/Rosetears520/aili-workflows --skill rose-memory

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

ROSE agents need reliable, project-local memory that preserves checkpoints, user requirements, and durable findings with evidence—without polluting global OpenCode config or losing context after interruptions.

Core Features & Use Cases

  • Project-local durable memory: Stores everything in memory/memory.db to keep state scoped to the current project.
  • Checkpoint and receipt tracking: Records task checkpoints and completion receipts to support safe resume and progress continuity.
  • Requirement-focused recall: Captures user-stated requirements, preferences, decisions, corrections, and acceptance criteria for reduced drift.
  • Evidence-backed durable findings: Promotes reusable project facts and lessons only when backed by evidence or an explicit no-promotion receipt.
  • Focused retrieval packs: Produces compact “pack” contexts (including task-focused packs) to recover relevant memory after compaction or ambiguity.

Quick Start

In your project root, initialize memory with the command: mkdir -p memory && rose-memory init --db memory/memory.db.

Frequently Asked Questions about rose-memory

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I persist task checkpoints and evidence for AI agent workflows?

You can persist task checkpoints and evidence by recording them in a project-local SQLite database. This approach stores state directly in your project directory to ensure progress continuity and safe resume after interruptions.

What is the best way to recover agent context after a session interruption?

Recovering agent context after a session interruption is done by requesting a focused retrieval pack from the SQLite memory store. This compact pack reconstructs the relevant context from previously recorded checkpoints and durable findings.

How do I store user requirements and project findings without polluting global config?

Storing user requirements and project findings without polluting global config requires a project-local memory database. This keeps state scoped to the current project directory rather than writing to global OpenCode configuration files.

Does the memory database enforce rules for promoting durable findings?

Yes, the memory database enforces strict additive rules for promoting durable findings. Reusable project facts and lessons are only recorded when they are backed by evidence or an explicit no-promotion receipt.

How do I initialize project-local memory in my workspace?

To initialize project-local memory, create a memory directory and run the init command with the database path. This sets up the SQLite database file to begin recording checkpoints and requirements.

When should I not use a project-local SQLite memory for agent state?

You should not use project-local SQLite memory if you need cross-project shared state or global configuration. The system deliberately scopes all checkpoints and evidence to the current project to prevent context drift and global pollution.