fleet-doctrine

What This Skill Does

The fleet-doctrine skill acts as the centralized intelligence and routing architecture for your OpenClaw agent fleet. Instead of relying on a single model for every interaction, this skill provides a structured decision-making framework to match specific task requirements with the strengths of the current top-tier AI models. It governs the orchestration of agents by defining roles for Opus (Command), Codex (Engineering), Sonnet (Operations), Gemini (Context/Vision), and Grok (Parallel/Validation). By implementing this doctrine, you ensure high-stakes tasks receive maximum reasoning capacity while routine, repeatable, and high-volume tasks remain cost-effective and efficient.

Installation

To integrate this routing strategy into your OpenClaw environment, execute the following command in your terminal: clawhub install openclaw/skills/skills/jarvis-drakon/fleet-doctrine

Use Cases

• Orchestrating complex project lifecycles by delegating sub-tasks to specialized models.
• Automating routine cron-based reports using lightweight, efficient models.
• Managing massive code refactoring projects where Codex handles logic and Grok handles sanity checks.
• Processing multi-modal data or massive documents exceeding 1M context limits using Gemini.
• Ensuring executive-level security and financial decisions are reviewed by Opus.

Example Prompts

1. "Route the repo refactor to the Chief Engineer role and ask Grok to perform a parallel sanity check on the security implications."
2. "I have an email summary task that needs to run every morning; configure this to use the Sonnet workhorse model."
3. "Analyze this 500-page regulatory document for potential compliance risks; use Gemini for its high-context capacity."

Tips & Limitations

• Intent over Rigidity: The doctrine is a guide for intent. If your primary model is unavailable, always prioritize task completion using available resources over total failure.
• Avoid Over-Spawning: A common anti-pattern is spawning multiple agents for simple tasks. Only use multi-model collaboration when you explicitly require a second opinion or cross-verification.
• Keep It Simple: Do not waste Opus's high-reasoning capabilities on templated form letters or simple email summaries.
• Maintenance: Regularly review your routing logic as new model versions are released, as the specific endpoints defined in the doctrine may need periodic updates to maintain optimal performance.