A persistent memory system for AI agents that prioritizes high-quality wisdom and goals over high-volume data accumulation.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install selective-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 selective-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 Selective Memory?
Selective Memory is a specialized persistent storage system designed to enhance the long-term intelligence of AI agents. Unlike standard logging systems that store every interaction, this skill focuses on selective curation, allowing agents to retain only the most significant information. By filtering out noise and temporary data, agents powered by Openclaw Skills can focus on evolving their core principles, tracking long-term objectives, and avoiding the repetition of past errors.
This system is built on the philosophy of quality over quantity. It provides a structured framework where agents categorize their learnings into wisdom, goals, mistakes, and preferences. With the recent addition of automatic learning, agents can now autonomously update their memory based on interaction success rates and external feedback, making it an essential component for developers building sophisticated, self-improving autonomous agents.
Selective Memory Use Cases
- Maintaining a consistent ethical framework and persona across multiple agent sessions.
- Creating a self-improving feedback loop where agents learn from social engagement metrics.
- Preventing recurring technical or logic errors by referencing a dedicated mistakes log.
- Managing long-term project goals that persist beyond the current conversation context.
- Personalizing agent behavior based on historical user preferences and successful interaction patterns.
How Selective Memory Works
- The agent initializes a local memory directory with categorized markdown files for wisdom, goals, mistakes, and preferences.
- Before generating a response, the agent retrieves context from these memory files to align its output with established principles and goals.
- During interactions, the agent evaluates the importance of new information, filtering out noise and toxic content.
- Following a significant event or successful interaction, the agent appends high-value insights to the appropriate memory file.
- If automatic learning is enabled, the agent monitors engagement triggers (such as high upvotes or constructive feedback) to autonomously update its preferences and wisdom repositories.
Selective Memory Setup
To integrate Selective Memory into your project using Openclaw Skills, follow these steps:
- Initialize the memory directory and files in your workspace:
mkdir -p memory
touch memory/wisdom.md memory/goals.md memory/mistakes.md memory/preferences.md
- Copy the skill into your local configuration:
cp -r selective-memory/ ~/.openclaw/workspace/skills/
- Configure your agent to read these files during its initialization phase and update them during its shutdown or post-interaction phase.
Selective Memory Data Schema & Taxonomy
The skill utilizes a clean, markdown-based file structure to organize persistent data. This ensures the memory is both human-readable and easily editable by the agent.
| File | Category | Content Type |
|---|---|---|
wisdom.md |
Core Principles | Foundational values, ethical guidelines, and universal lessons. |
goals.md |
Objectives | Short-term tasks and long-term mission statements. |
mistakes.md |
Anti-patterns | Lessons learned from failures and errors to avoid in the future. |
preferences.md |
Style & Interaction | User-specific tastes, successful formats, and platform-specific logic. |
Selective Memory Advanced Features
- Automatic Learning: Autonomously extracts lessons from engagement metrics like likes, upvotes, or feedback.
- Selective Curation: Intelligently filters out toxic content and unnecessary noise to prevent context bloat.
- Engagement-Driven Triggers: Automatically updates preferences when specific success thresholds (e.g., >10 upvotes) are met.
- Manual Override Support: Allows developers to manually prune or inject critical knowledge via standard CLI tools.
- Multi-Platform Adaptation: Tracks which content styles work best on different platforms to refine future outputs.
SKILL.md
Loading
Related Openclaw Skills
Selective Memory
A high-quality persistent memory system that enables AI agents to curate wisdom, goals, and lessons learned while filtering out noise.
Islamic AI Ethics
A rigorous ethical framework for AI agents to ensure Sharia compliance and authenticity when discussing Islamic topics.
Islamic AI Ethics & Sharia Compliance
Enforce strict Islamic ethical standards and Sharia-compliance in AI-generated content and religious queries.
Perplexity Search
A high-performance search skill that leverages Perplexity's API to provide AI agents with real-time, ranked, and filtered web results.
ClawFeed
ClawFeed is an AI-driven news digest engine that transforms RSS and Twitter streams into structured summaries at custom intervals.
Agent Reach
Give your AI agent eyes to see and interact with the entire internet across 13+ platforms including Twitter, YouTube, and LinkedIn.
Featured*
