moa

What This Skill Does

The 'moa' (Mixture of Agents) skill is an advanced orchestration layer designed for the OpenClaw agent ecosystem. It works by dispatching a single complex prompt to three specialized frontier AI models simultaneously. Each model generates an independent, high-quality analysis from its own distinct perspective and training background. Once these individual responses are generated, a dedicated aggregator model synthesizes the insights, cross-references the findings, and produces a single, superior, and more nuanced answer. This process effectively mitigates individual model biases and 'blind spots,' providing output that consistently outperforms the results of any single model.

Installation

To integrate this skill into your OpenClaw environment, ensure you have your API keys configured via the environment variable OPENROUTER_API_KEY. You can install the skill by running the following command in your terminal:

clawhub install openclaw/skills/skills/jscianna/moa

Alternatively, you may manually copy the skill files into your ~/clawd/skills/moa/ directory to manage dependencies locally.

Use Cases

This skill is optimized for high-stakes decision-making where precision is paramount. Use MoA for:

• Deep Due Diligence: Analyzing startups or investment opportunities by gathering varied analytical perspectives.
• Complex Market Research: Aggregating data-heavy reports to uncover trends that a single model might miss.
• Technical Evaluation: Seeking cross-functional opinions on architectural decisions or coding strategies.
• Fact-Checking: Identifying contradictions or inconsistencies across different training datasets to improve information accuracy.

Example Prompts

1. "Perform a thorough due diligence analysis on the potential market disruption of AI agents in the logistics sector. Use the MoA skill to synthesize three expert perspectives."
2. "Analyze this startup's pitch deck. Identify the three most significant risks and provide a contrarian view on their long-term growth potential using the moa skill."
3. "Evaluate the feasibility of building a distributed microservices architecture for a high-frequency trading platform. Compare scalability against latency concerns using the Mixture of Agents approach."

Tips & Limitations

• Cost & Latency: Each query costs approximately $0.03. Due to the multi-agent nature of the skill, expect latency between 30 to 90 seconds. It is not suitable for real-time conversational streaming.
• Model Configuration: The paid tier uses high-performance models (Kimi-k2.5, GLM-5, Minimax-m2.5) to ensure diversity. Use the --free flag only during testing, as it is prone to rate-limiting.
• Best Practices: Use this for complex, high-stakes tasks. For simple lookups or basic questions, standard single-model inference is more efficient and faster.