model-tiering

Model tiering for AI agents standardizes operational capacity by assigning a single tier per role, preventing both under- and over-provisioning. It applies across CEO, chiefs, and specialists, enforcing governance rules to match model capacity with project needs.

You assign model tiers by providing a new agent name and triggering the tiering policy, which applies the hierarchical tier map to the agent's frontmatter. This enforces governance rules and ensures consistent tiering across the program.

Model tiering requires parsing agent frontmatter from agents/*.md files and applying the hierarchical tier map. You need agents defined with frontmatter to ensure consistent tiering and maintain the audit trail of overrides across the program.

Model tiering for agents differs from manual assignment by enforcing a hierarchical tier policy that automatically applies governance rules and upgrade or downgrade criteria. This standardization prevents both under- and over-provisioning while maintaining traceable overrides.