gov-model-selection-governance

What problem does it solve?

Provide a consistent, auditable way to select, document, and maintain model metadata so agent and skill packages use task-fit models and avoid broken integrations or unsafe model assumptions across providers.

Core Features & Use Cases

• Canonical profile assignment: Choose one of the five repo-wide profiles (deep-reasoning, balanced-execution, guided-workflow, utility-ops, routing-inherit) to ensure consistent routing and cost/control tradeoffs.
• Provider-safe mappings: Populate provider_models with preferred and availability-safe defaults, including recommended fields for cursor and vscode-copilot to avoid blank fallbacks.
• Governance checks & tooling: Commands and audit checks to find missing model_profile entries, orphaned profile names, and to document rationale when non-obvious choices are made.
• Use case: Onboarding a new skill package and assigning it a balanced-execution profile with mapped fallbacks for Claude Code, Codex, Cursor, Copilot, and ChatGPT.

Quick Start

Run the gov-model-selection-governance skill to assign a canonical model_profile and provider mappings to a package manifest and record the selection rationale.