skill

claude-advisor-plugin

Claude Code plugin: Advisor Strategy — Opus reasons and reviews while Sonnet subagents handle implementation.

zabrodsk/claude-advisor-plugincommunitylistedunreviewed

Set up

Skills are installed by placing their folder where your agent loads skills.

Download ZIP from GitHub Open the repository

The download comes from github.com/zabrodsk/claude-advisor-plugin itself, not from our server, so what you get is exactly what its author published.

Install instructions come from the README

Skills do not use an MCP config. Follow the steps in the README below.

README

synced Oct 23, 2026

Claude Opus is the smartest model Anthropic ships. It's also the most expensive. The Advisor Strategy lets you pay for Opus reasoning once and get parallel Sonnet execution for everything else. Same decisions. Roughly a third of the tokens.

Cost savings, in practice

Approximate, directional comparison for a typical session — not a measured benchmark.

Install

Inside Claude Code, run these two commands separately — send the first, wait for it to succeed, then send the second:

1. Add the marketplace:

/plugin marketplace add https://github.com/zabrodsk/claude-advisor-plugin.git

2. Install the plugin:

/plugin install advisor@claude-advisor-plugin

Pasting both lines in one prompt will concatenate the second command onto the clone URL and fail. Send each line on its own.

Usage

/claude-advisor-plugin:advisor

Or say "advisor mode", "use advisor", or "advisor strategy" — the skill auto-triggers.

How it works

Opus dispatches executors via the Agent tool with model: "sonnet". If three rounds haven't resolved the task, Opus surfaces the blocker rather than burning more tokens.

Without advisor	With advisor
Opus writes every file edit	Sonnet writes; Opus reviews
Sequential execution	Parallel Sonnet subagents
Opus tokens for boilerplate	Opus tokens for reasoning only
No spend cap	Hard cap: 3 executor rounds

Why you want this

Opus decisions, Sonnet bills. Deep reasoning where it matters, fast implementation where it doesn't.
Parallel by default. Independent subtasks run simultaneously, not one after another.
Hard cap on spend. Three executor rounds maximum — no silent runaway costs.
One command to activate. Drop-in for any session, works alongside any other plugin or workflow.

Inspired by Anthropic's Advisor Strategy — the same cost/quality tradeoff as the advisor_20260301 API tool, implemented at the orchestration layer.