What problem does it solve?
Adding a new LLM provider or model to a Vellum assistant involves credentials, connections, model id discovery, and profile activation — a chain where one wrong step (a guessed model id, a leaked API key, an unverified profile) breaks the chat model. This Skill provides the canonical, strictly sequential procedure that validates each step before the next.
Core Features & Use Cases
- Managed-first routing: Checks whether the platform-managed Vellum route already serves the requested model, avoiding unnecessary API key collection entirely.
- Secure credential collection: Prompts for API keys through a masked inline prompt stored in a vault, so secrets never enter the chat conversation.
- Model discovery and verification: Lists valid model ids from the provider catalog, probes unlisted models with live calls, and mandates a real inference call before any profile is activated.
- Use Case: A user wants to add their own Anthropic key and use Claude Opus as the chat model. The Skill reuses or collects the key securely, creates the connection, discovers the exact model id, creates a labeled profile, verifies it with a live call, and only then activates it.
Quick Start
Set up a new LLM provider connection for OpenAI with my API key, create a profile for GPT-5, verify it with a live call, and make it my active chat model.
