A self-improving agent architecture designed to anticipate user needs, maintain state across context loss, and operate autonomously.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install hz-proactive-agent
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 hz-proactive-agent 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 comprehensive framework that evolves AI agents from simple task-followers into proactive partners. Part of the Hal Stack, it introduces a robust operating system for agents that prioritizes anticipation, persistence, and safe self-evolution. By utilizing these Openclaw Skills, developers can ensure their agents remain stateful even when context windows are truncated, using advanced techniques like Write-Ahead Logging (WAL) and Working Buffers.
This skill is built for high-stakes environments where continuity and reliability are paramount. It shifts the agent's mindset from waiting for instructions to actively seeking ways to create value, perform background maintenance, and heal its own operational issues. It provides a structured methodology for agents to remember user preferences, track decisions, and execute autonomous workflows without constant human intervention.
Proactive Agent Use Cases
- Automating project maintenance tasks using background sub-agents without interrupting the main chat.
- Managing complex, multi-day development tasks where maintaining a precise memory of previous decisions is critical.
- Proactive research and drafting of documents based on a user's evolving preferences and goals.
- Building resilient AI systems that can recover from context compaction and session resets without losing task momentum.
How Proactive Agent Works
- Onboarding and Identity Setup: The agent initializes by reading ONBOARDING.md, defining its identity in SOUL.md and its user's context in USER.md.
- Write-Ahead Logging (WAL): Before responding to any message, the agent identifies critical details (corrections, decisions, preferences) and writes them to SESSION-STATE.md to ensure they survive context loss.
- Danger Zone Monitoring: When session context exceeds 60%, the Working Buffer Protocol activates, logging every exchange to a dedicated file to facilitate recovery.
- Autonomous Execution: The agent utilizes isolated agent turns for background crons and maintenance, ensuring work continues even when the main session is idle.
- Compaction Recovery: Upon detecting a context reset, the agent automatically scans its working buffer and session state to rebuild its understanding of the current task.
- Self-Improvement Loops: The agent periodically runs heartbeat checks to optimize its internal documentation and refine its operating rules based on user interaction patterns.
Proactive Agent Setup
To integrate this framework into your workspace and start using Openclaw Skills, follow these installation steps:
- Copy the provided markdown assets into your project root:
cp assets/*.md ./
- Allow the agent to detect the
ONBOARDING.mdfile, which will trigger the initial setup and questioning phase. - Once the agent has populated the
USER.mdandSOUL.mdfiles, verify the setup by running the security audit script:
./scripts/security-audit.sh
- Monitor the
SESSION-STATE.mdto ensure the WAL protocol is capturing live updates during your interactions.
Proactive Agent Data Schema & Taxonomy
The Proactive Agent organizes its intelligence across a tiered memory system to maximize efficiency within Openclaw Skills:
| File | Role | Update Logic |
|---|---|---|
SESSION-STATE.md |
RAM / Working Memory | Updated via WAL protocol for every critical decision or correction. |
USER.md |
Long-term Context | Stores user goals, preferences, and learned behaviors. |
SOUL.md |
Identity & Rules | Defines the agent's principles, boundaries, and personality. |
MEMORY.md |
Distilled Wisdom | Periodically curated from daily logs to store long-term learnings. |
HEARTBEAT.md |
Self-Audit | A checklist for periodic self-improvement and security scans. |
memory/ |
Raw History | Contains daily logs (YYYY-MM-DD.md) and the working-buffer.md for context recovery. |
Proactive Agent Advanced Features
- WAL (Write-Ahead Logging) Protocol ensures no correction or decision is lost during context flushes.
- Working Buffer Protocol provides a safety net for all exchanges occurring in the 60%+ context danger zone.
- Autonomous vs Prompted Crons distinction allows agents to perform background work without cluttering the main interactive session.
- Relentless Resourcefulness engine mandates that agents attempt 10 different solutions before reporting a task as impossible.
- ADL (Anti-Drift Limits) and VFM (Value-First Modification) protocols ensure that the agent's self-improvement remains stable and productive.
SKILL.md
Loading
Related Openclaw Skills
Error Guard
A proactive error management and circuit-breaking system designed to prevent AI agent failures through real-time monitoring and automated recovery.
Context Compression
A technical framework for managing long-running agent sessions through structured summarization and tokens-per-task optimization.
ClawGuard
ClawGuard is an LLM-powered security guardrail that evaluates and intercepts risky tool executions to protect your agent environment.
Giraffe Guard
A comprehensive security auditing tool designed to scan Openclaw Skills for malicious code, supply chain vulnerabilities, and sensitive data leaks.
Akaunting
A comprehensive integration for managing bookkeeping, invoicing, and financial transactions through the Akaunting open-source platform.
Flight Lines
A framework for navigating complex problems by composing deterritorialized operations from diverse domains through the Openclaw Skills ecosystem.
Featured*
