Memory Defragmenter

Defragment and optimize agent memory files by cleaning duplicates, merging similar entries, archiving stale content, and ensuring proper tiering. Use when: (1) memory files feel cluttered or bloated; (2) before or after context optimization; (3) weekly memory maintenance; (4) when explicitly asked to clean, defragment, or optimize memory.

Clean, merge, and optimize memory files to keep them lean and effective.

Purpose

Memory files grow over time and become:

• Bloated with duplicates
• Stale with outdated info
• Scattered across tiers
• Inconsistent in format

This skill defragments them.

When to Run

Trigger	Action
Weekly maintenance	Full defragment
Before context optimization	Clean first
After many sessions	Remove stale entries
Explicit request	Clean specified files

Defragmentation Workflow

Step 1: Analyze Memory State

bash

python3 scripts/analyze_memory.py

Reports:

• Files analyzed and sizes
• Duplicate detection
• Stale entry detection (30+ days old)
• Tier distribution
• Formatting consistency

Step 2: Generate Defragmentation Plan

bash

python3 scripts/defragment.py --plan

Creates defragment-plan.md with:

• Entries to merge
• Entries to archive
• Entries to delete
• Entries to promote/demote

Step 3: Review Plan

Read the plan and approve/modify before execution.

Step 4: Execute Defragmentation

bash

python3 scripts/defragment.py --execute

Backs up original files first, then:

• Merges duplicate entries
• Archives stale content to archive/
• Promotes hot entries to HOT tier
• Demotes cold entries to archive
• Normalizes formatting

Step 5: Verify

bash

python3 scripts/verify_memory.py

Checks:

• All files readable
• No broken links
• Consistent formatting
• Tier limits respected

Memory Tiers Reference

Tier	Location	Max Size	Age Policy
HOT	memory.md, ~/self-improving/memory.md	≤100 lines each	Review monthly
WARM	memory/*.md, ~/self-improving/domains/	≤200 lines each	Review quarterly
COLD	archive/	Unlimited	Archive at 90 days

Safety Rules

1. Always backup before executing
2. Review plan before applying
3. Never delete - only archive
4. Keep originals for 24h after
5. Verify after every defragment

Files

• scripts/analyze_memory.py — Analyze memory state
• scripts/defragment.py — Plan and execute cleanup
• scripts/verify_memory.py — Verify integrity
• references/rules.md — Defragmentation rules and patterns