tier-down

Implement tier-down policies and token ceilings for Model Router cost controls.

Updated May 11, 2026

Search Tags:#caching #cost-optimization #harness #token-management #ai-infrastructure #model-router

Ready to run

npx skills add https://github.com/AesopScott/mojo --skill tier-down

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: tier-down
Source: https://github.com/AesopScott/mojo/tree/main/harnesses/skills/tier-down
Command: npx skills add https://github.com/AesopScott/mojo --skill tier-down

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill addresses excessive operational costs and inefficient token usage within the Model Router harness by enforcing strict routing policies and caching strategies.

Core Features & Use Cases

Cost-Efficient Routing: Automatically directs low-complexity tasks to cheaper model tiers.
Token Ceiling Enforcement: Sets hard limits on per-route token consumption to prevent runaway costs.
Cache-First Dispatch: Prioritizes existing cache hits before triggering expensive external API calls.
Use Case: When managing a high-volume AI application, use this skill to implement a policy that routes simple classification tasks to a smaller, faster model while reserving high-end models for complex reasoning.

Quick Start

Use the tier-down skill to analyze the current Model Router configuration and propose a new token ceiling policy for the cheap-task route.

Frequently Asked Questions about tier-down

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I reduce AI model routing costs and optimize token usage in my infrastructure?▼

To reduce AI model routing costs, you can enforce cost-efficiency controls and token usage limits within the Model Router harness. This approach manages spend, compute waste, and API call frequency to ensure operational stability.

What is cache-first dispatch for API calls and when should I use it?▼

Cache-first dispatch is a routing strategy that prioritizes existing cache hits before triggering expensive external API calls. Use it in high-volume AI applications to automatically direct low-complexity tasks to cheaper model tiers.

How do I set per-route token ceilings to prevent runaway AI costs?▼

You set per-route token ceilings by analyzing your Model Router configuration and proposing a new token-down policy. This enforces hard limits on token consumption for specific routes, preventing runaway costs.

Can I route simple classification tasks to a smaller model while reserving high-end models for complex reasoning?▼

Yes, you can route simple classification tasks to a smaller, faster model while reserving high-end models for complex reasoning. This cost-efficient routing is achieved by configuring tier-down policies within the Model Router harness.

Does the tier-down skill require any specific dependencies or components to manage API call frequency?▼

No, the tier-down skill requires no specific dependencies or components to manage API call frequency. It operates within the Model Router harness to enforce strict routing policies and caching strategies natively.