multi-llm-integration

What problem does it solve?

Integrating specialized external LLMs into a Claude-centered development pipeline can be fragmented and error-prone, forcing teams to manually call APIs, manage keys, and merge outputs into project specs. This skill provides a clear architecture and practical guidance to route tasks to the best-suited model (Gemini, GPT, Mistral) and reliably integrate their outputs into the specs/ driven workflow.

Core Features & Use Cases

• Provider orchestration: Maps tasks to recommended models (e.g., Gemini for visual/UX, GPT for in-depth analysis, Mistral for cost-efficient translation) and documents endpoint and auth patterns.
• Script-driven integration: Describes scripts that call external APIs, accept llm-config.json, and push results into specs/ for downstream agents and /dev-sync ingestion.
• Config and security guidance: Shows environment variable usage and a local llm-config.json pattern to centralize API keys and provider selection.
• Use case: Generate UX wireframes with Gemini, run GPT-based market analysis for a new feature, and merge both outputs into specs/03-user-stories.md for review by the Claude agents.

Quick Start

Run the generate-ux.ts script with your configured llm-config.json to produce UX wireframes for the "User Dashboard" feature and then run the /dev-sync process to merge the outputs into the project's specs.