A cognitive workflow skill that enables AI agents to evaluate complex problems through multiple distinct personas before synthesizing a final solution.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install diverge
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 diverge 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 Diverge?
Diverge is a sophisticated architecture for AI agents that prevents narrow thinking by forcing the evaluation of a problem from multiple angles. Instead of providing a single immediate response, the agent identifies when a situation requires diverse viewpoints—such as technical, legal, or user-centric perspectives—and simulates an internal panel of experts. This approach is fundamental to Openclaw Skills that aim for high-reliability outputs in complex domains.
By systematically exploring various directions before converging, Diverge ensures that blind spots are addressed and trade-offs are explicitly documented. This skill transforms a standard LLM interaction into a robust analytical process, making it ideal for high-stakes environments where a single viewpoint is insufficient.
Diverge Use Cases
- Product management decisions where power users and new users have conflicting requirements.
- Safety-critical evaluations requiring distinct viewpoints from doctors, lawyers, and security experts.
- Creative brainstorming sessions that need multiple aesthetic directions before selecting a final concept.
- Complex technical architecture reviews where security, performance, and maintainability must be balanced.
How Diverge Works
- The agent evaluates the task complexity to determine if a divergent approach is necessary.
- It selects the appropriate number of perspectives (ranging from 2 for low complexity to 7 for high complexity).
- Relevant personas are chosen from the personas.md template library based on the problem domain.
- Each perspective generates an independent analysis, highlighting specific risks and benefits.
- The agent synthesizes the findings by identifying conflicts, common ground, and weighing priorities.
- A final decision is documented, clearly outlining the chosen path and the trade-offs involved.
Diverge Setup
To integrate this capability into your agent environment using Openclaw Skills, follow these steps:
# Navigate to your agent's skill directory
cd .ai/skills
# Clone or add the Diverge skill
openclaw add diverge
# Configure your perspective templates
nano diverge/personas.md
Diverge Data Schema & Taxonomy
The Diverge skill organizes its logic and configuration through structured Markdown files and tracking metadata:
| Component | Description |
|---|---|
SKILL.md |
Core logic defining when and how to spawn multiple perspectives. |
personas.md |
A library of expert templates (e.g., Security, Legal, Product) for the agent to adopt. |
synthesis.md |
Contains conflict resolution patterns and weighting logic for the convergence phase. |
Learning Logs |
Internal metadata tracking user preference for speed vs. thoroughness to adapt future behavior. |
Diverge Advanced Features
- Dynamic Scaling: Automatically adjusts the quantity of perspectives (2-7) based on real-time task complexity analysis.
- User Preference Adaptation: Learns over time whether a specific user values rapid responses or deep, thorough exploration.
- Stake-Weighted Synthesis: Built-in logic to prioritize safety and regulatory concerns over general aesthetic or speed preferences.
- Multi-Agent Orchestration: Compatible with other Openclaw Skills like delegate to route specific perspectives to specialized sub-agents.
SKILL.md
Loading
Related Openclaw Skills
Digital Marketing
A comprehensive growth marketing system for planning, launching, and optimizing revenue-focused digital campaigns and funnel operations.
Dietitian
A technical AI skill for structured meal planning, precise calorie targeting, and macro-nutrient optimization based on physiological data.
Diagram
An AI-powered visualization tool for generating technical diagrams, system architectures, and workflows using Mermaid, PlantUML, and ASCII.
Design System
A powerful tool for building scalable design systems using token-first architecture and consistent component patterns.
DNS Configuration Skill
A specialized skill for configuring high-performance, secure, and reliable DNS records with expert-level migration and authentication strategies.
Docker Compose
Define and manage multi-container applications with precise dependency handling, networking, and volume management.
Featured*
