start-gpowers-multimodel

Activates the GPowers Multi-Model Orchestrator persona with per-role subagent model routing.

Updated Jul 7, 2026
One-click install
npx skills add https://github.com/ricardolui/gcp-custom-agent-skills --skill start-gpowers-multimodel-ricardolui
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: start-gpowers-multimodel
Source: https://github.com/ricardolui/gcp-custom-agent-skills/tree/main/start-gpowers-multimodel
Command: npx skills add https://github.com/ricardolui/gcp-custom-agent-skills --skill start-gpowers-multimodel-ricardolui

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Coordinating multi-agent coding workflows often wastes cost or quality by using one model for every role. This Skill activates the GPowers Multi-Model Orchestrator persona, which routes subagent tasks to appropriate models: fast flash models for execution and verification, and inherited high-capability models for architecture, planning, and lead review. ## Core Features & Use Cases - Persona Activation: Adopts or delegates to the gpowers_multimodel orchestrator via explicit commands like /start-gpowers-multimodel or @start-gpowers-multimodel. - Dynamic Model Routing: Enforces a routing matrix where implementers and verifiers use Model: "flash" while architects, planners, and lead reviewers use Model: "inherit". - Tiered Pipelines: Applies three orchestration tiers, from single-file flash edits to parallel Best-of-N architectural pipelines with mixed-model reviewers. - Use Case: A developer asks to refactor a cross-cutting module; the orchestrator spawns an architect and two planners on inherited models, an implementer and verifier on flash, and two reviewers split across both model tiers. ## Quick Start Activate the GPowers Multi-Model Orchestrator by typing /start-gpowers-multimodel and then describe the coding task you want orchestrated.

Frequently Asked Questions about start-gpowers-multimodel

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

FAQPage Schema
How do I activate the GPowers Multi-Model Orchestrator?

Activate it by typing /start-gpowers-multimodel, @start-gpowers-multimodel, or /gpowers_multimodel, or by asking to use gpowers_multimodel. For native routing in the main chat, select gpowers_multimodel from the Agent Selector dropdown in the Jetski UI.

How does multi-model routing assign models to subagents?

Routing assigns Model: "flash" to implementers, verifiers, and routine reviewers, and Model: "inherit" to architects, planners, and lead reviewers. Every entry in the invoke_subagent Subagents array must include an explicit Model value.

What are the orchestration tiers in gpowers_multimodel?

Tier 1 handles single-file edits with one flash implementer. Tier 2 runs a sequential planner-implementer-reviewer-verifier pipeline on flash. Tier 3 runs a parallel Best-of-N pipeline mixing inherit and flash models for architectural work.

Why is per-role model routing not working in my session?

Per-role routing requires always_inherit_model set to false in the invoke_subagent tool configuration. If you are running as the default main agent, switch to gpowers_multimodel in the Agent Selector or delegate the task via invoke_subagent.

Can the orchestrator edit code directly in the main context?

No. The orchestrator never executes implementation edits, bug fixes, or test runs in the main context. It delegates 100% of execution to subagents using the structured [ROLE: <agent_name>] handoff envelope.