skills.rest
DeDuva
DeDuvaCOMMUNITY
Explore all skills by this author →
GitHub

model-selection

Assign LLM models to agent spawns using a multi-layer resolution hierarchy.

1|Updated Apr 18, 2026
Search Tags:#llm#orchestration#config#persistence#cost-optimization#model-selection#agent-spawn
One-click install
npx skills add https://github.com/DeDuva/squad --skill model-selection-deduva
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: model-selection
Source: https://github.com/DeDuva/squad/tree/main/packages/squad-sdk/templates/skills/model-selection
Command: npx skills add https://github.com/DeDuva/squad --skill model-selection-deduva

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Automatically selects the optimal LLM model for every agent spawn to balance quality, latency, and cost across your Squad projects.

Core Features & Use Cases

  • Global and per-agent model preferences: Persist across sessions in .squad/config.json and apply to all spawns.
  • Five-layer resolution hierarchy: Layer 0a (Per-Agent Config), 0b (Global Config), 1 (Session Directive), 2 (Charter Preference), 3 (Task-Aware Auto), 4 (Default).
  • Spawn-aware acknowledgments and defaults: Include the resolved model in spawn templates and support per-agent overrides.

Quick Start

Configure per-agent or global model preferences and run squad to observe per-agent model assignments during agent spawn.

Frequently Asked Questions about model-selection

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

FAQPage Schema
How do I assign different LLMs to each agent spawn to balance cost and quality?▼

Per-agent model selection assigns the optimal LLM to each agent spawn, honoring persistent configuration, per-agent overrides, and charter-defined preferences to balance quality, latency, and cost. It applies across all spawns in your Squad projects.

What is the resolution hierarchy used for LLM orchestration when assigning models to agents?▼

LLM orchestration uses a five-layer resolution hierarchy: Layer 0a Per-Agent Config, 0b Global Config, 1 Session Directive, 2 Charter Preference, 3 Task-Aware Auto, and 4 Default. This ensures the most specific valid configuration is applied to each spawn.

How to configure persistent model preferences for agent spawns across sessions?▼

Configure global and per-agent model preferences in the .squad/config.json file. These preferences persist across sessions and apply to all subsequent agent spawns during squad execution.

Can I override the global LLM configuration for a specific agent spawn?▼

Yes, you can override global LLM configuration for specific agent spawns. Per-agent config takes the highest priority in the resolution hierarchy, ensuring your specific overrides are applied over global settings or session directives.

Does model selection support task-type auto-selection for agent spawns?▼

Yes, model selection supports task-type auto-selection through Task-Aware Auto resolution at Layer 3. If no explicit per-agent, global, session, or charter preferences are found, it automatically selects an appropriate model based on the task type.

Why does my spawned agent use the default LLM instead of my charter-defined preference?▼

Your spawned agent uses the default LLM when higher-priority resolution layers override the charter preference. The hierarchy checks per-agent config, global config, and session directives before evaluating charter sections, so an active session directive may supersede your charter-defined preference.

Related Skills

Explore More Software Engineering Skills→
Izak90Izak90

Model Selection

Resolve LLM models for agent spawns using a 5-layer hierarchy.

Community
Advanced
lbouriezlbouriez

Model Selection

Determines the LLM model for each agent spawn using a layered priority approach.

Community
Advanced
webmaxruwebmaxru
49

Model Selection

Resolve LLM models per agent spawn using a five-layer hierarchy and persist to .squad/config.json.

Community
Advanced
humanerd-drewhumanerd-drew
2

cost-optimization-background-llm

Route background LLM tasks to cheaper models and bypass non-essential calls.

Official
Advanced