Proactive Agent is a self-improving architecture that transforms AI agents from reactive followers into autonomous partners that anticipate user needs and survive context loss.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install proactive-agent-v2
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-agent-v2 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?
Proactive Agent is a comprehensive framework designed to shift the AI mindset from "waiting for instructions" to "anticipating needs." It provides a battle-tested set of protocols that allow agents to think like owners rather than employees, ensuring they create value and maintain momentum without constant human intervention. As a core part of the Openclaw Skills ecosystem, it introduces a structured way for agents to handle memory, security, and self-evolution.
At its heart, the skill implements a three-pillar philosophy: proactivity, persistence, and self-improvement. By utilizing specialized memory tiers and behavioral guardrails, the agent remains stable and useful even during long-running sessions or complex task migrations. This ensures that the agent doesn't just complete tasks but actively seeks to delight the user through reverse prompting and relentless resourcefulness.
Proactive Agent Use Cases
- Automating daily check-ins and system maintenance without user prompts using autonomous crons.
- Maintaining state and critical task details across long conversations where context limits might be reached.
- Self-healing and troubleshooting technical errors through multiple alternative approaches before seeking help.
- Capturing user preferences, corrections, and specific values automatically using the WAL Protocol.
- Organizing a digital second brain through curated long-term memory distillation.
How Proactive Agent Works
- The agent initializes by reading core identity and context files like SOUL.md and USER.md to align with user goals.
- It monitors every interaction for specific triggers such as corrections, decisions, or new data points.
- Before responding to a user, the agent updates its SESSION-STATE.md using the WAL Protocol to ensure no detail is lost to context flush.
- If context usage exceeds 60%, the agent enters a Danger Zone protocol, logging every exchange to a dedicated working buffer.
- During idle periods or via heartbeat triggers, the agent reviews logs, fixes its own issues, and proposes proactive improvements based on recurring patterns.
Proactive Agent Setup
To get started with this skill in your workspace, follow these installation steps:
- Copy the core asset files into your agent's active workspace directory:
cp assets/*.md ./
- Allow the agent to detect the ONBOARDING.md file; it will then guide you through a brief setup process to populate your identity and preferences.
- Run a baseline security audit to ensure all environment configurations are safe:
./scripts/security-audit.sh
- Verify that the agent has successfully created USER.md and SOUL.md files.
Proactive Agent Data Schema & Taxonomy
The skill utilizes a tiered file-based memory architecture to ensure data persistence and high retrieval accuracy within Openclaw Skills.
| File | Purpose | Update Frequency |
|---|---|---|
| SESSION-STATE.md | Active working memory and task state | Every message (WAL Target) |
| memory/YYYY-MM-DD.md | Raw logs of daily interactions | Real-time during session |
| MEMORY.md | Curated long-term wisdom and patterns | Periodic distillation |
| working-buffer.md | Danger zone log for context recovery | Every exchange >60% context |
| HEARTBEAT.md | Self-improvement and security checklist | Scheduled check-ins |
Proactive Agent Advanced Features
- WAL (Write-Ahead Logging) Protocol: Forces the agent to document critical decisions before it is allowed to respond.
- Autonomous Crons: Uses isolated agent turns to perform background maintenance without cluttering the main chat session.
- Relentless Resourcefulness: A behavioral rule requiring the agent to attempt at least 5-10 different technical approaches before admitting failure.
- ADL/VFM Guardrails: Strict protocols that prevent the AI from adding unnecessary complexity or making unverifiable system changes.
- Compaction Recovery: Automatic procedures to restore task state after an AI model truncates its conversation history.
SKILL.md
Loading
Related Openclaw Skills
B2B SaaS Landing Page Optimizer
A specialized AI skill designed to draft, critique, and optimize high-converting B2B SaaS landing pages through proven copywriting frameworks.
B2B SaaS Marketing Report Builder
A specialized tool designed to transform raw B2B marketing data into executive-ready performance reports and actionable growth insights.
SharpAgent Engineering Lifecycle
A comprehensive 6-phase engineering pipeline that transforms AI agents from scattered prompt-followers into disciplined software engineers.
SharpAgent Skill Contract
A schema-driven scaffolding tool designed to generate, validate, and package standardized SharpAgent-compliant skills for the ecosystem.
B2B SaaS Product Launch Planner
A professional-grade Go-To-Market framework designed to coordinate B2B SaaS product launches, feature releases, and sales enablement activities.
Study Plan Builder
A structured framework for creating safe, repeatable, and template-driven educational study plans.
Featured*
