A production-grade governance and memory optimization system designed to transform basic AI agents into high-performing, self-improving swarms.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install clawhub-memory-tiers-pro
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install clawhub-memory-tiers-pro using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Agent Memory Tiers Pro?
Agent Memory Tiers Pro is an advanced extension for the foundational memory system, specifically engineered for developers managing multi-agent swarms in production environments. It moves beyond simple data persistence by introducing a rigorous 8-point quality grading rubric, progressive disclosure patterns for extreme token efficiency, and formalized orchestrator protocols.
Developed from real-world experience running a 20-agent production swarm, this skill provides the structural integrity needed to scale AI operations. It ensures that every agent operates within defined boundaries, handles errors gracefully, and contributes to a collective self-improvement loop through centralized lessons and pattern documentation within Openclaw Skills.
Agent Memory Tiers Pro Use Cases
- Managing a large-scale swarm of 10+ agents where coordination and task routing are critical.
- Optimizing token usage in high-frequency agent activations through progressive disclosure.
- Auditing and improving agent performance using a standardized, objective scoring rubric.
- Establishing automated handoff protocols between specialist agents to ensure data integrity.
- Building a self-healing AI system that logs mistakes and adapts workflows over time.
How Agent Memory Tiers Pro Works
- The system implements a 4-tier Progressive Disclosure model, loading only essential identity and context files (L0/L1) initially to save tokens.
- Agents are evaluated against an 8-point rubric (Role Clarity, Triggers, Workflow, etc.) to ensure they meet production standards.
- An Orchestrator Agent utilizes the status check protocol to monitor swarm health and route tasks based on agent availability and flags.
- Cross-agent communication is handled via write-once handoff files in a shared workspace to prevent data corruption.
- Continuous improvement is driven by a Monthly Swarm Audit that analyzes the lessons.md and patterns.md files to refine agent SOUL.md instructions.
Agent Memory Tiers Pro Setup
To use this professional tier, you must first have the base memory skill installed. Follow these steps to initialize the pro features in your Openclaw Skills environment:
# Ensure the base skill is active
claw install agent-memory-tiers
# Install the Pro extension
claw install agent-memory-tiers-pro
# Initialize the tracking files in your main workspace
touch workspace/lessons.md workspace/patterns.md
mkdir -p workspace/scorecards workspace/handoffs
Agent Memory Tiers Pro Data Schema & Taxonomy
The skill organizes data across a tiered hierarchy to balance accessibility with token economy:
| Tier | File / Location | Content Type | Retention |
|---|---|---|---|
| T1 | L0.md |
Identity & Status Flag | Every Run |
| T2 | L1.md |
7-day Rolling State | Every Run |
| T3 | SOUL.md |
Core Instructions | On Activation |
| T4 | references/ |
Large Data & Docs | On Demand |
| Logs | lessons.md |
Error & Fix History | Permanent |
| Rules | patterns.md |
Proven Best Practices | Permanent |
Agent Memory Tiers Pro Advanced Features
- 8-Point Quality Rubric: Standardized A/C/F grading system for auditing agent reliability.
- Orchestrator Protocol: Dedicated logic for task routing, swarm status monitoring, and escalation management.
- Token Budgeting: Strict SOUL.md line-count limits (target < 300 lines) with external reference pointers.
- Self-Improvement Loop: Automated mechanisms for agents to record and learn from operational failures.
- Handoff Management: Formalized filesystem-based messaging for reliable inter-agent data transfers.
SKILL.md
Loading
Related Openclaw Skills
Evidence Organizer
A specialized tool designed to categorize, structure, and catalog evidence for legal proceedings and discovery.
SEO Expert
An advanced SEO and Generative Engine Optimization (GEO) knowledge base designed to guide AI agents through professional search engine growth strategies.
Garbage Classification
A comprehensive AI skill for accurate waste sorting, professional recycling services, and environmental education.
Fractional CFO Playbook
A comprehensive operational framework for managing outsourced finance executive services and client deliverables.
fxtwitter-translate
A specialized tool for fetching full-text content from Twitter/X links via the fxtwitter API and generating high-quality Chinese translations in markdown.
Ops
A comprehensive operational framework for managing incidents, deployments, team rituals, and vendor relationships to ensure organizational reliability.
Featured*
