skills.rest
KimYx0207
KimYx0207COMMUNITY
Explore all skills by this author →
𝕏@KimYx0207GitHub

memory-3layer

Build a three-layer memory system storing facts, notes, and tacit knowledge across sessions.

160|49|Updated Feb 14, 2026
Search Tags:#claude code#hooks#codex#migration#agent#memory#session state
One-click install
npx skills add https://github.com/KimYx0207/Kim_Service --skill memory-3layer
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: memory-3layer
Source: https://github.com/KimYx0207/Kim_Service/tree/main/skills/memory-3layer
Command: npx skills add https://github.com/KimYx0207/Kim_Service --skill memory-3layer

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) and assets (resource) components.

What problem does it solve?

This Skill gives long-running Agent work a durable, inspectable memory layer so important facts, daily progress, and lasting lessons are not lost between sessions. It is designed to replace ad hoc notes and unsafe full-transcript retention with a structured, reviewable memory system.

Core Features & Use Cases

  • Three-layer memory model: Stores structured facts in Layer 1, daily notes in Layer 2, and durable tacit knowledge in Layer 3.
  • Cross-session continuity: Loads only bounded, relevant memory so new sessions can resume context without flooding the prompt.
  • Legacy migration and review: Migrates old .claude/memory data, checks for conflicts or duplicates, and supports explicit promotion to superseded status.
  • Platform-neutral operation: Works with Claude Code and Codex through Hooks, while still supporting a manual core for other hosts.
  • Use case: A team can keep project decisions, recurring implementation details, and daily progress synchronized across multiple agent sessions without relying on raw chat logs.

Quick Start

Ask the Skill to initialize or load the project memory, then record only confirmed facts or migrate legacy .claude/memory data when you need a clean, platform-neutral memory tree.

Frequently Asked Questions about memory-3layer

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

FAQPage Schema
How do I keep agent context and session state across multiple Claude Code sessions?▼

To keep agent context across multiple Claude Code sessions, you need a structured memory system that stores durable facts, daily notes, and tacit knowledge in project-local directories for bounded context loading. This preserves important details without flooding the prompt.

What is the best way to migrate legacy .claude/memory data without losing facts?▼

Migrating legacy .claude/memory data safely requires a non-destructive process that checks for conflicts or duplicates, supports explicit promotion to superseded status, and filters sensitive data while preserving the original records during the transition.

Does this agent memory system work with Codex and manual hosts, or only Claude Code?▼

This agent memory system works with Codex and Claude Code through safe hook integration, while also supporting a manual core for other hosts. It operates platform-neutrally, requiring only project-local memory directories and Python 3.8+.

How do I structure long-running agent memory to avoid unbounded prompt context?▼

Structuring long-running agent memory to avoid unbounded prompt context involves using a three-layer model: Layer 1 for structured facts, Layer 2 for daily notes, and Layer 3 for durable tacit knowledge, loading only relevant bounded context per session.

Do I need Python installed to use hooks for cross-session agent memory?▼

Yes, you need Python 3.8 or higher installed to use this cross-session agent memory system. It relies on Python scripts for platform-neutral operation, lifecycle status tracking, and safe hook integration with Claude Code or Codex.

Why does my agent lose project decisions and recurring implementation details between sessions?▼

Agents lose project decisions between sessions when relying on ad hoc notes or raw chat logs instead of a structured memory layer. A three-layer system explicitly records confirmed facts and daily progress to ensure cross-session continuity.

Related Skills

Explore More Software Engineering Skills→
aaron-he-zhuaaron-he-zhu
142

memory-management

Manage two-layer memory with hot cache and cold storage for SEO projects.

Community
Basic
whanyu1212whanyu1212
29

memory

Load MEMORY.md and log events to HISTORY.md for persistent AI context.

Community
Intermediate
Undertone0809Undertone0809
280

para-memory-files

Organize persistent agent memory into YAML facts and markdown daily notes.

Community
Advanced
jkeith10jkeith10

para-memory-files

Organize and persist knowledge across sessions using the PARA memory model.

Community
Advanced