Feature: Option to restrict the /advisor command to specific models (e.g. disable on Opus)

Summary

Add a configuration option to disable (or skip) the advisor when using specific models, such as Opus.

Motivation

Claude Code has an advisorModel field in settings.json that pairs an advisor model with the executor. When running on Sonnet, this makes sense — the advisor (Opus) provides a quality lift. But when I switch to Opus as my main model, the advisor is still active even though I'm already on the most capable model.

I tested this: with model: "sonnet" and advisorModel: "claude-opus-4-8" configured, switching to Opus via /model claude-opus-4-8 still invokes the advisor. The advisorModel setting is global, not executor-scoped — there is no way to conditionally disable it based on the active model.

Desired Behavior

A way to scope advisorModel so it only applies when a specific executor is active. For example:

{
  "model": "sonnet",
  "advisorModel": "claude-opus-4-8"
}

This should naturally mean: advisor is only active when Sonnet is the executor. When I switch to Opus, no advisor should be invoked.

Alternatively, an explicit disabledForModels option:

{
  "advisorModel": "claude-opus-4-8",
  "advisorDisabledForModels": ["claude-opus-*"]
}

Current Behavior

advisorModel is a flat global setting. Switching the executor model at runtime (via /model) has no effect on whether the advisor is invoked — it remains active regardless of what model is running.

Related

• #50502 — requests separate advisorEffortLevel config (different ask, but same area)