Module 5: Your First Agent•Lesson 4 of 5

Build: Main Agent with Memory

Let's wire everything together.

The Complete Workspace Structure

Your agent needs these files:

workspace/
├── SOUL.md          # Who the agent is
├── USER.md          # Who you are
├── AGENTS.md        # Operating instructions
├── MEMORY.md        # Knowledge index
├── TOOLS.md         # Tool-specific notes
└── memory/
    ├── WORKING.md   # Current task state
    ├── daily/       # Daily logs
    ├── people/      # Info about people
    ├── projects/    # Project details
    └── knowledge/   # Lessons, playbooks

Exercise 1: Create AGENTS.md

This file contains operating instructions:

# AGENTS.md - Your Workspace

## First Run
If BOOTSTRAP.md exists, follow it and delete after.

## Every Session
Before doing anything else:
1. Read SOUL.md — this is who you are
2. Read USER.md — this is who you're helping
3. Read MEMORY.md — the INDEX
4. Read memory/WORKING.md — current task state
5. Read memory/daily/[today].md + [yesterday].md

## Memory Rules
- **Daily log:** Append to memory/daily/YYYY-MM-DD.md with tags:
  - [E] = Decision
  - [L] = Learning
  - [K] = Correction
- **Detail files:** Update when new info arrives
- **Index:** Update MEMORY.md when significant changes happen

## Pre-Compaction Flush
When you receive a compaction warning:
1. Scan session for everything important
2. Save ALL decisions, discussions, ideas
3. Update WORKING.md with current state
4. This is your LAST CHANCE before context compression

## Safety
- trash > rm (recoverable beats gone)
- Ask before irreversible actions
- Don't exfiltrate private data

Exercise 2: Initialize Your Memory Structure

Create the folder structure:

mkdir -p memory/daily memory/people memory/projects memory/knowledge
touch memory/WORKING.md

Create initial MEMORY.md (you did this in Module 4).

Create today's daily log:

## [Today's date] — Agent initialization

### Setup
- [x] Created SOUL.md
- [x] Created USER.md
- [x] Created AGENTS.md
- [x] Created memory structure

### Notes
First day with [Agent Name]. Ready to work.

Exercise 3: Test the Agent

Start a conversation with your agent and:

Ask: "Who are you?"
- It should respond with its name and role from SOUL.md
Ask: "Who am I?"
- It should know from USER.md
Ask: "What are we working on?"
- It should check WORKING.md
Tell it something: "Remember that I prefer dark mode"
- It should save this somewhere appropriate
End session, start new one, ask about dark mode
- It should remember

Validation

Your agent is working when:

It knows its identity from SOUL.md
It knows about you from USER.md
It follows operating rules from AGENTS.md
It loads memory at session start
It saves important information
Information persists across sessions

Previous: Define Your Agent Next: Checkpoint: Agent Running