A lightweight memory management skill that periodically reloads agent identity and distills logs into actionable long-term context.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install agents-refresh
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 agents-refresh 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 Refresh?
The Agent Memory Refresh skill is designed to solve the common problem of personality drift and context loss in long-running AI sessions. By implementing a sophisticated 3-layer memory system, it ensures your agent remains rational and focused on its core identity. This skill is a vital addition to the ecosystem of Openclaw Skills, providing a mechanism to keep the agent soul, user preferences, and project goals fresh in the model active context without needing to start new sessions.
Built specifically for lightweight distributions, this version relies on Markdown-based configuration and manual or scheduled tool calls rather than external scripts. It empowers developers to build more reliable and human-like agents that can handle multi-day projects with ease, making it a cornerstone for anyone building advanced automation via Openclaw Skills.
Agent Memory Refresh Use Cases
- Maintaining agent personality and soul during extensive coding or research sessions.
- Preventing context hallucinations by periodically reloading immutable core identity files.
- Distilling daily activity logs into high-level memory summaries for better long-term performance.
- Automating project status updates through scheduled heartbeat checks within Openclaw Skills.
How Agent Memory Refresh Works
- The agent monitors the context length and session duration using system tools like session_status.
- It triggers a refresh cycle manually or via a cron job to read the AGENTS.md, IDENTITY.md, and SOUL.md files.
- The Gold Mining process begins, where the agent extracts decisions, tasks, lessons, and seeds from current logs.
- Extracted insights are distilled into the MEMORY.md file, updating the active ToDo list and project foundation.
- The agent re-synchronizes its internal state with the refreshed 3-layer memory structure to ensure continuity.
Agent Memory Refresh Setup
To integrate this into your environment for Openclaw Skills, follow these steps:
- Modify Heartbeat: Add the refresh task to your HEARTBEAT.md file:
- [ ] Refresh: read workspace/{AGENTS.md,IDENTITY.md,SOUL.md,USER.md} -> Chain SOUL/USER/IDENTITY + Summarize MEMORY
- Configure Cron: Execute the following command to schedule a daily refresh at 6 AM:
cron action=add job='{"name":"agents-refresh-daily","schedule":{"kind":"cron","expr":"0 6 * * *","tz":"Asia/Tokyo"},"payload":{"kind":"systemEvent","text":"[Daily Refresh] read workspace/{AGENTS.md,IDENTITY.md,SOUL.md,USER.md}! Reload SOUL/USER/IDENTITY chain -> Keep focus on ToDo/projects! Summarize recent MEMORY/ToDo."},"delivery":{"mode":"announce"},"sessionTarget":"main","task":"refresh_memory"}'
Agent Memory Refresh Data Schema & Taxonomy
The skill organizes information across a structured hierarchy to optimize Openclaw Skills performance:
| File | Type | Purpose |
|---|---|---|
| IDENTITY.md | Immutable | The agent core soul, name, vibe, and hard rules. |
| MEMORY.md | Dynamic | Current progress, key decisions, and prioritized ToDo lists. |
| memory/YYYY-MM-DD.md | Log | Historical archive of daily chats and actions. |
| HEARTBEAT.md | Control | Task list for triggering periodic maintenance routines. |
Agent Memory Refresh Advanced Features
- Gold Mining Distillation: Automatically separates Confirmed Matters from Idea Seeds to keep the active memory lean.
- Cron Integration: Supports complex scheduling like 9AM/1PM/8PM rotations for continuous context updates.
- Zero-Dependency Architecture: Functions entirely through Markdown and standard tool calls, compatible with all Openclaw Skills environments.
- Context Length Monitoring: Ability to trigger refreshes based on session_status to proactively prevent memory saturation.
SKILL.md
Loading
Related Openclaw Skills
Skill Hunter
A powerful discovery and security vetting engine designed to help developers navigate and install over 10,000 available skills on ClawHub.
Cheat Code
Cheat Code expands your AI agent's capabilities by providing real-time synthesis and data from outside its static training set.
Create Skill
A comprehensive framework for building, importing, and registering modular capabilities to enhance AI agent performance and domain expertise.
Bot Customizer
An automated utility to personalize AI agent personas and instructions by merging custom user requirements into existing agent configurations.
voiceclaw-jp
A voice-first interface for OpenClaw that enables real-time spoken conversations using wake words, streaming LLM responses, and VOICEVOX text-to-speech.
AI News Collector
A high-performance AI agent skill for multi-dimensional search, aggregation, and ranking of the latest news and trends in the artificial intelligence sector.
Featured*
