Smart Multi-Model Routing & Delegation

Subagent Orchestration is a high-level routing skill designed for developers who use multiple AI models (Claude, GPT/Codex, and Gemini) and want to optimize for cost, performance, and context window limitations. Instead of manually switching browser tabs or CLI tools, this skill allows your primary agent to intelligently delegate sub-tasks to the best-fitting model for the job.

How it works

The skill acts as an intelligent traffic controller. It analyzes your request and routes it to specific vendor CLIs based on task strengths: Claude for precise code refactoring, Gemini for massive context ingestion (up to 1M+ tokens), or Codex for web-integrated research. It handles the syntax differences between each model's CLI, enabling seamless cross-model execution.

Why use this skill?

• Cost Management: Automatically route lightweight formatting or summary tasks to cheaper models like Gemini Flash.
• Rate Limit Resilience: If Claude is rate-limited, the skill can automatically failover to Gemini or Codex to keep your workflow moving.
• Cross-Model Verification: Run the same security or logic check across all three models simultaneously to find "blind spots" through a side-by-side comparison table.
• Context Optimization: Automatically detects when a prompt exceeds standard context limits and routes it to the large-window models.

Supported Integration

The skill leverages your existing local environment, supporting official CLIs for Claude, Codex/ChatGPT, and Gemini. It outputs clean, prefixed text for single runs or structured Markdown tables for cross-model comparisons.