Memory Structure

Enable AI to learn from mistakes and never repeat them. Error tracking, layered memory, regular self-reflection, continuous improvement.

Skill: memory-structure

Name

memory-structure

Function

Copy existing memory structure to new environments, enabling other Agents to use the same memory system.

Use Cases

• Quickly set up self-improvement framework for new Agents
• Unify memory organization across multiple Agents
• Establish standardized self-reflection process

Skill Contents

This skill includes:

File	Description
memory.md	Main memory file: preferences, patterns, rules (HOT tier)
corrections.md	Error correction log: "what I got wrong" and correct answers
index.md	Memory index: tracks all memory file updates
heartbeat-state.md	Heartbeat state: records self-reflection check timestamps and results
heartbeat-rules.md	Heartbeat rules: defines triggers and execution logic for self-reflection
setup.md	Setup guide: installation and configuration instructions

Usage

1. Install Skill

Using ClawHub:

bash

clawhub install memory-structure

2. Initialize Memory Structure

Create memory directory in target workspace:

bash

mkdir -p ~/self-improving/{domains,projects,archive}

Then manually create the following files (copy template contents below):

3. Regular Self-Reflection

After completing important tasks, Agent should update heartbeat-state.md with reflection results.

Core Concepts

• HOT Tier: Memory files are hot-tier, high-frequency access
• Corrections Log: Errors are the source of improvement, must be recorded
• Heartbeat: Regular self-reflection to check if memory structure needs updates

Dependencies

• No external dependencies
• Requires file system write permission
• Heartbeat interval configured in OpenClaw config (default: 30 minutes)