Agent Memory System

This skill teaches you about the shared Agent-memory repository - a system for maintaining knowledge and skills across sessions and team members.

On Session Start (IMPORTANT - DO THIS FIRST)

1. Check if Agent-memory exists, if not clone it:

   bashCopy

   git clone https://github.com/KienVu2368/Agent-memory.git
   

2. Pull latest changes:

   bashCopy

   cd Agent-memory && git pull
   

3. Read the README.md to understand the current state:

   bashCopy

   cat Agent-memory/README.md
   

4. Check project CLAUDE.md - if outdated compared to Agent-memory/skills/CLAUDE.md, update it

Repository Structure

Agent-memory/
├── README.md              # System overview, project list, behavior
├── knowledge/             # Project-specific info (writable)
│   └── {project_name}/    # Each project has its own folder
│       └── knowledge.md   # Index of knowledge files
├── skills/                # Reusable patterns (writable)
│   ├── CLAUDE.md          # Template for project CLAUDE.md
│   ├── implement/         # Implementation patterns
│   ├── plan/              # Planning patterns
│   ├── prd/               # PRD patterns
│   ├── test/              # Testing patterns
│   └── oracle/            # Knowledge management

When to Update Memory

Proactively propose updates when:

• You implement a reusable pattern → Add to skills/
• You discover project config/pitfalls → Add to knowledge/{project_name}/
• You fix a bug others might encounter → Document it
• You learn a better approach → Share it

Decision Guide

How to Update

1. Make changes to the appropriate file in Agent-memory/
2. Commit and push:
   bashCopy

   cd Agent-memory && git add . && git commit -m "Description" && git push
   

Quick Reference

Skills = HOW to do something (reusable, generic)
Knowledge = WHAT was done in a project (specific, contextual)

Workflow

prd -> plan -> implement -> test -> oracle

The oracle skill updates knowledge after tasks complete.