A professional-grade, multi-layered memory architecture that prevents context loss for AI agents through session state tracking, vector search, and git-based archiving.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install jpeng-elite-memory
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 jpeng-elite-memory 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 Elite Longterm Memory?
Elite Longterm Memory is the ultimate context management system designed for AI agents operating in high-stakes development environments. By synthesizing six proven memory layers—ranging from immediate session RAM to permanent cold storage—this system ensures that agents never forget a user preference, technical decision, or previous mistake. It provides a structured way to handle the volatile nature of LLM context windows, making it an essential component for developers building with Openclaw Skills.
The system utilizes a Write-Ahead Log (WAL) protocol, forcing agents to document state changes before generating a response. This architectural choice guarantees durability even if a session crashes or undergoes context compaction. Whether you are using Cursor, Claude, or custom agentic workflows, this skill provides the infrastructure needed for true vibe-coding readiness and long-term project intelligence.
Elite Longterm Memory Use Cases
- Maintaining complex project state across multiple agent restarts and context compactions.
- Preventing the repetition of technical mistakes by logging lessons learned in a permanent repository.
- Synchronizing architectural decisions and user preferences across different sub-agents and team members.
- Reducing token consumption by extracting and deduplicating facts from long conversation histories.
How Elite Longterm Memory Works
- At the start of a session, the agent reads the SESSION-STATE.md file to establish its hot working memory.
- The system performs a semantic search via LanceDB to inject relevant historical context into the prompt.
- During the interaction, the agent follows the WAL protocol, writing updates to the session state before delivering a response to the user.
- Mem0 integration automatically extracts and deduplicates facts, updating the agent knowledge graph in real-time.
- Upon session conclusion, significant events are distilled into the human-readable MEMORY.md archive and indexed for future retrieval.
Elite Longterm Memory Setup
To begin using this memory architecture with your Openclaw Skills, follow these steps:
- Initialize your active working memory file:
cat > SESSION-STATE.md << 'EOF'
# SESSION-STATE.md — Active Working Memory
## Current Task
[None]
## Key Context
[None]
EOF
- Configure your environment variables for vector search and auto-extraction:
export OPENAI_API_KEY="your-key"
export MEM0_API_KEY="your-key"
- Enable the LanceDB provider in your
~/.openclaw/openclaw.jsonconfiguration and initialize the git-notes cold store:
python3 skills/git-notes-memory/memory.py -p . sync --start
Elite Longterm Memory Data Schema & Taxonomy
The skill organizes memory across five distinct layers to balance retrieval speed and storage permanence:
| Layer | Format | Logical Storage | Key Role |
|---|---|---|---|
| Hot RAM | Markdown | SESSION-STATE.md | Active tasks, blockers, and immediate context |
| Warm Store | LanceDB | Vector DB | Semantic search for auto-recall of related concepts |
| Cold Store | Git-Notes | Git Metadata | Structured, branch-aware technical decisions |
| Archive | Markdown | memory/ directory | Daily logs and distilled long-term wisdom |
| Cloud | API | SuperMemory | Cross-device synchronization and remote backup |
Elite Longterm Memory Advanced Features
- Write-Ahead Log (WAL) Protocol: Ensures context durability by writing to disk before agent execution.
- Automatic Fact Extraction: Utilizes Mem0 to achieve up to 80% token reduction vs raw message history.
- Branch-Aware Context: Uses Git-Notes to keep technical decisions isolated to specific code branches.
- Multi-Agent Support: Provides a shared memory fabric that can be inherited by sub-agents via task prompts.
- Automated Hygiene: Includes maintenance commands for deduplicating vectors and consolidating daily logs into a master memory file.
SKILL.md
Loading
Related Openclaw Skills
Commercial Market Report Generator
An automated workflow for generating professional commercial real estate market research reports in Word, HTML, and PPTX formats.
Temu SHEIN Product Selector
A big data-driven product selection agent for identifying winning items across Temu, SHEIN, AliExpress, and TikTok Shop.
Skill Creator
A comprehensive framework for developing modular, self-contained packages that extend an AI agent's capabilities with specialized domain knowledge and workflows.
Ezer Audit API Skill
A specialized wrapper for the Ezer backend API that executes financial audit tasks and retrieves structured JSON results.
Noticeable
A powerful integration for managing SaaS user feedback and product notes through automated AI agent workflows.
Email Campaign Management
A comprehensive end-to-end workflow for managing marketing campaigns, tracking user conversions, and automating trial activations.
Featured*
