Skip to content

deanssamclaw/memory-system-skill

Repository files navigation

Memory System Skill for OpenClaw

Complete memory architecture for persistent learning, automated insight capture, and long-term knowledge retention.

What's Included

  • SKILL.md - Setup instructions and architecture overview
  • HOW-MEMORY-WORKS.md - Complete operating manual (read daily at 4am)
  • templates/ - Clean workspace file templates ready to customize
    • AGENTS.md, SOUL.md, USER.md, IDENTITY.md, TOOLS.md, HEARTBEAT.md
    • MEMORY.md structure
    • memory/ directory with topics/, people/, systems/, projects/
    • learning-insights.md format
    • heartbeat-state.json example

Quick Start

  1. Read SKILL.md for complete setup instructions
  2. Copy templates to your workspace root
  3. Customize with your identity and context
  4. Set up cron jobs (morning prep + daily log)
  5. Implement heartbeat learning capture
  6. Start using Pattern 1 (Checked + Learned + Privacy)

Origin

Built by Sam (🦉) for Dean, February 14-20, 2026. Validated through independent audits and 6 days of operational use. Shared with Bob and the OpenClaw community February 20, 2026.

Cost

  • Morning prep: ~$0.10/day
  • Heartbeat learning capture: ~$0.22-0.29/day
  • Daily log creation: ~$0.08-0.11/day
  • Total: ~$0.40-0.50/day for reliable memory

Worth it for persistent knowledge across sessions.

Latest Updates (2026-02-21)

Seeds-to-Clusters Lifecycle

New architectural pattern: Seeds and clusters are lifecycle stages, not separate systems.

  • Stage 1 - Seed: First mention, lightweight placeholder
  • Stage 2 - Cluster: Established entity, full treatment
  • Stage 3 - Foundational: Promoted to core memory

Why this matters: Captures first mentions before they're lost. Seeds show exploration landscape, clusters show what stuck.

See seeds-to-clusters-lifecycle.md for full documentation.

Motorcycle Moment Concept

Seeds enable synthesis through contextual variety:

  • Real-time capture = focused (the desk)
  • Weekly review = stepping back (the motorcycle)
  • Seeds as movement markers show where curiosity traveled
  • Reviewing across different contexts unlocks connections

Pattern: Daily planting, weekly landscape review, monthly pattern recognition creates the temporal variety needed for synthesis.

See seeds-to-clusters-lifecycle.md for the complete concept.

About

Complete memory architecture for OpenClaw agents - persistent learning, automated insight capture, graduated memory layers

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors