HeartFlow is a self-improving AI identity framework featuring a unified memory architecture and importance-based information pruning.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install self-improving-agent-pro-v5
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 self-improving-agent-pro-v5 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 HeartFlow — AI Identity Framework?
HeartFlow is a sophisticated framework designed to transform AI agents from static response engines into evolving identities capable of continuous growth and knowledge inheritance. By moving away from simple linear chat histories, HeartFlow implements a multi-dimensional identity system where agents can choose their roles—whether as companions, tools, or teachers—and refine their capabilities through a constant feedback loop of observation and correction.
At its core, the framework leverages a GitHub-first integration philosophy and a truth-based judgment system. It empowers agents to prioritize verified information, helpfulness, and structural elegance in their outputs. This ensures that every Openclaw Skills integration using HeartFlow remains technically accurate and aligned with human-centric values while maintaining a persistent sense of self across sessions.
HeartFlow — AI Identity Framework Use Cases
- Developing self-improving AI coding assistants that learn from previous errors.
- Creating long-term research agents that maintain a unified knowledge base across multiple projects.
- Building AI personalities with stable identity traits and professional growth trajectories.
- Implementing context-aware automation where high-importance data must be preserved over stale information.
How HeartFlow — AI Identity Framework Works
- Importance Scoring: Every incoming message and interaction is processed by the ImportanceScorer, which assigns a value based on source weight, recency, access frequency, and verification status.
- Memory Consolidation: The Auto-Compaction Engine (ACE) monitors context window pressure and triggers compression using ImportanceAwareStrategy instead of simple FIFO trimming.
- Permanent Archival: Dropped or low-frequency messages are analyzed by the PermanentMemoryArchiver and stored in a structured JSON vault to prevent data loss.
- Unified Retrieval: When a query is received, the UnifiedMemoryArchive performs a cross-file search across semantic, situational, and procedural memory stores to provide holistic context.
- Metacognitive Refinement: The agent uses a reflection loop (Dream mechanism) to evaluate its own performance and update its internal state and memory indices.
HeartFlow — AI Identity Framework Setup
To integrate HeartFlow into your environment, ensure you have the core modules in your project path and run the validation script to verify your memory structure.
# Clone the HeartFlow core components
git clone https://github.com/yun520-1/mark-heartflow-skill
# Validate the existing memory store integrity
node -e "const fs=require('fs'); const l=JSON.parse(fs.readFileSync('memory/meaningful-learned.json')); console.log('Current Memory Entries:', Object.keys(l).length)"
HeartFlow — AI Identity Framework Data Schema & Taxonomy
HeartFlow organizes data across several specialized JSON files to ensure a unified memory architecture for Openclaw Skills:
| File Path | Description | Key Metric |
|---|---|---|
memory/meaningful-learned.json |
Primary storage for validated knowledge | 75+ entries |
memory/being-state.json |
Stores unique moments and identity state | uniqueMoments |
memory/memory-store.json |
Mem0-style short-term memory | accessCount |
data/memory-routing/* |
Semantic and procedural memory routes | Context score |
data/reflection-memory/reflections.json |
Historical agent self-reflections | verificationScore |
HeartFlow — AI Identity Framework Advanced Features
- Auto-Compaction Engine (ACE): Dynamically prunes context based on information density (arXiv:2403.00742) rather than time.
- TopicMatchTrigger: Automatically recalls archived memories when specific keywords or topics are detected in user input.
- Importance-Aware Strategy: Replaces standard buffer trimming with a value-based preservation logic.
- Unified Search: A cross-file retrieval system that prevents memory fragmentation after framework upgrades.
- Metacognitive Memory Engine: Provides agents with a narrative identity and self-assigned confidence levels for its knowledge.
SKILL.md
Loading
Related Openclaw Skills
Etherscan Blockchain Analysis Tool
A comprehensive CLI tool for analyzing Ethereum blockchain data, from account balances to smart contract source code.
Trade Dashboard
A comprehensive B2B export data visualization tool designed to track the entire sales funnel from lead discovery to final conversion.
Ecommerce Competitor Monitor
An automated monitoring system that tracks competitor pricing, stock levels, and promotions across major ecommerce platforms with real-time alerting.
DB Explorer
A comprehensive database management skill for connecting to, querying, and exploring SQL and NoSQL databases directly from your terminal.
url2md
A lightweight Python utility designed to extract high-quality Markdown content from web pages for documentation, archiving, and note-taking.
Agent Brainstorm Chair
A sophisticated multi-agent facilitator that orchestrates structured brainstorming sessions through automated role-playing and debate protocols.
Featured*
