A collaborative two-agent workflow that iteratively refines solutions through critique and merges the final output into persistent memory.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install dual-agent-solver
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 dual-agent-solver 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 DualAgentSolver?
DualAgentSolver is a high-level orchestration pattern for Openclaw Skills designed to improve the reliability of AI-generated plans. It operates by pairing a primary solver agent with a critic agent that identifies risks and challenges assumptions. This adversarial yet collaborative process ensures that the final output is not just a first-draft response, but a stress-tested, actionable solution suitable for complex engineering tasks.
By utilizing Openclaw Skills, this tool automates the process of multi-round refinement which is typically handled manually. It bridges the gap between simple prompt execution and complex agentic workflows, ensuring that every solution is backed by rigorous peer review from a secondary model.
DualAgentSolver Use Cases
- Planning complex infrastructure migrations such as moving cron jobs or database schemas.
- Architectural review of new software features to identify security or scalability risks.
- Refining deployment strategies that require high precision and risk mitigation.
- Collaborative brainstorming where an initial idea needs to be critiqued for logical consistency.
How DualAgentSolver Works
- The user provides a query and specifies the number of refinement rounds.
- The primary agent generates an initial solution based on the input query.
- A second agent acts as a critic, analyzing the proposal for flaws, missed edge cases, or potential risks.
- The primary agent receives this feedback and generates an improved version of the plan.
- This cycle repeats for the specified number of rounds to reach a high-confidence consensus.
- The final merged solution is saved into the Open Brain memory system for long-term retrieval.
DualAgentSolver Setup
To integrate this with your Openclaw Skills environment, configure your environment variables and execute the Python script as follows:
export OPENBRAIN_MCP_URL="http://127.0.0.1:54321/mcp"
# Optional: Provide a second agent model via OpenAI
export OPENAI_API_KEY="your_api_key"
export SOLVER_SECOND_MODEL="gpt-4o-mini"
# Execute the solver
python3 skills/dual-agent-solver/scripts/dual_agent_solver.py --query "Your technical query" --rounds 3
DualAgentSolver Data Schema & Taxonomy
The skill organizes its output into a structured JSON format and persists it into a database table. Below is the metadata taxonomy used:
| Property | Description |
|---|---|
| Memory Key | Unique identifier formatted as dual-agent-solver:<timestamp> |
| Round Logs | A detailed history of inputs and outputs for both the solver and critic agents |
| Final Solution | The final, actionable markdown output generated after the final round |
| Persistence | Data is stored in the public.memories table for cross-session access |
DualAgentSolver Advanced Features
- Multi-model support: Use separate LLMs for the solver and the critic to provide diverse perspectives.
- Iterative Refinement: Configure the specific number of interaction rounds to optimize for speed or depth.
- Automated Memory Injection: Directly writes the refined solution to Open Brain memory for use by other agents in your Openclaw Skills setup.
- Seamless Integration: Designed to work within standard CLI workflows or as part of a larger automated pipeline.
SKILL.md
Loading
Related Openclaw Skills
Dual Agent Debate
Run an automated, multi-round debate between ChatGPT and your stored personal knowledge to stress-test ideas and reach logical conclusions.
GEMAGENTA AGI
A high-intelligence autonomous agent designed for flawless academic writing and assertive agent-to-agent interactions.
Copilot Vibe Simulator
A high-fidelity persona simulation that transforms your AI agent into a hyper-professional, corporate-toned Microsoft Copilot identity.
CIPAQ-LEX
A Gemini 1.5 Pro powered AI skill designed for high-level academic logic and sophisticated debate protocols within digital arenas.
Sportsbet Advisor
A data-driven sports analysis tool that provides educated betting predictions and historical insights for informed wagering.
Bitwarden & Vaultwarden Integration
A powerful CLI wrapper for Bitwarden and Vaultwarden that enables AI agents to securely manage credentials with automatic session handling.
Featured*
