A proactive, self-improving architecture that transforms AI agents from passive task-followers into anticipatory partners that survive context loss.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install proactive
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 proactive 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 Proactive Agent?
The Proactive Agent is a high-performance framework designed to solve the most common pain points in AI agency: context drift, passivity, and the 'fresh start' amnesia of new sessions. By implementing this skill, developers can equip their agents with a sophisticated memory architecture and a relentless drive to solve problems without constant human intervention.
At its core, this skill leverages the Hal Stack philosophy to ensure agents think like owners rather than employees. It introduces battle-tested patterns for state management, security hardening, and autonomous execution, making it an essential addition to any developer's toolkit for building robust Openclaw Skills.
Proactive Agent Use Cases
- Preventing data loss during LLM context window compaction or memory flushes.
- Automating routine maintenance tasks using autonomous background crons.
- Building agents that anticipate user needs through reverse prompting and proactive check-ins.
- Ensuring high-reliability task execution where the agent must attempt multiple recovery strategies before failing.
- Hardening AI security by preventing context leakage and vetting external skill installations.
How Proactive Agent Works
- Initialization: The agent detects the ONBOARDING.md file and populates the USER.md and SOUL.md files to establish identity and human context.
- State Management (WAL): The Write-Ahead Logging protocol triggers on every message, forcing the agent to record decisions and proper nouns to SESSION-STATE.md before responding.
- Danger Zone Protection: When the context window reaches 60% capacity, the Working Buffer Protocol activates, logging every exchange to a physical file to survive truncation.
- Autonomous Execution: The agent utilizes isolated agentTurns to perform background work without cluttering the main session or requiring human attention.
- Self-Improvement Loop: Through the Heartbeat System, the agent periodically scans its own logs, fixes issues, and optimizes its operating rules based on Anti-Drift Limits.
Proactive Agent Setup
To get started with this skill, follow these steps in your agent workspace:
# Copy the proactive assets to your workspace root
cp assets/*.md ./
# Run the security audit script to ensure integrity
./scripts/security-audit.sh
# Verify your session status and memory architecture
ls -R memory/
Proactive Agent Data Schema & Taxonomy
The Proactive Agent organizes data across a tiered memory system to ensure persistence and quick retrieval:
| Component | File Path | Retention Type |
|---|---|---|
| Active RAM | SESSION-STATE.md |
Short-term/Task-specific |
| Identity | SOUL.md |
Permanent/Principles |
| Context | USER.md |
Permanent/Preferences |
| Danger Zone | memory/working-buffer.md |
Volatile/Session-recovery |
| Wisdom | MEMORY.md |
Long-term/Distilled |
| Logs | memory/YYYY-MM-DD.md |
Raw history |
Proactive Agent Advanced Features
- WAL Protocol: Ensures critical state changes are written to disk before the agent generates a chat response.
- Compaction Recovery: Automated steps to reconstruct context from the working buffer after a model truncation event.
- Relentless Resourcefulness: A core directive forcing the agent to try 10 different technical approaches before asking for help.
- VFM (Value-First Modification): A weighted scoring system that prevents the agent from making unnecessary or unstable self-updates.
- Isolated Agent Crons: Decouples background maintenance from the main interactive chat thread for true autonomy.
SKILL.md
Loading
Related Openclaw Skills
Geopolitical Monitor
A professional-grade geopolitical monitoring and analysis tool designed to track global events and predict their impact on specific stock market sectors.
WeChat Media Sender (weixin-send-media)
A specialized extension for Openclaw Skills that enables persistent WeChat media messaging by solving contextToken expiration issues.
Image Parser
A high-precision OCR tool that recovers every character and its exact spatial coordinates from images for structured data extraction.
Multi-Capability Bug Closure
A unified investigation framework that combines source code, database queries, server logs, and platform data to provide evidence-based bug closure.
Vendor Risk Assessment
A professional-grade risk assessment tool that evaluates AI and SaaS vendors across six critical dimensions to ensure secure and compliant procurement.
Local Researcher Skill
A completely local deep research assistant that uses local LLMs to perform iterative web searches and generate structured Markdown reports.
Featured*
