Discover available local or low-cost models (Llama, Qwen, Mistral, etc., via Ollama/LM Studio/LMDeploy, plus cloud fallbacks). Decide which model is best suited to a task (Q&A, planning, tool-use, coding, summarization, vision, etc.). Route and “wire in” the right model at runtime, minimizing Claude/OpenAI usage while preserving capabilities.
This skill includes a SKILL.md with agent instructions.