Intelligent Multi-Model Delegation

In high-stakes development, no single AI model is perfect for every task. Gemini excels at architectural reasoning, Grok dominates fast CLI execution, and Codex is superior for structured patching. This skill transforms your AI agent into a sophisticated project manager capable of orchestrating these diverse strengths into a unified workflow.

What it does

The Orchestrator analyzes complex technical goals and decomposes them into a phased execution plan. It cross-references your specific task requirements against a shared "Consensus Technical Mapping Table" to ensure every sub-task is routed to the model best equipped to handle it.

• Architectural Mapping: Uses Gemini's massive context window to map monorepos and define intent.
• Phase Decomposition: Breaks monolithic prompts into actionable search, read, edit, and plan stages.
• Smart Handoffs: Recommends precise "Agent Relays" including validation steps to ensure data integrity between model switches.

Why use this skill

Generic prompting often hits context limits or suffers from "model hallucination" when a task drifts outside a specific model's comfort zone. This skill eliminates the guesswork of deciding which model to use. It provides a structured "Agent Relay" plan that leverages shared schemas and metadata validators, ensuring that when you switch from Gemini to Codex, the context remains consistent and the output remains high-quality.