Advisor Skills for Claude Code

Source on GitHub

Advisor Skills lets Claude Code consult other models and agents. When a senior engineer gets stuck, they ask for a second opinion or go read the docs; these skills give Claude the same options.

What This Plugin Does

  1. Break Tunnel Vision: The Strategist agent gives a high-level outside view on architectural decisions. It runs on Opus as a sub-agent, so you only pay for the expensive model on the decisions that need it.

  2. Verify & Research: Gemini fetches current documentation and library info, so Claude doesn’t hallucinate outdated APIs.

  3. Deepen Reasoning: Codex takes algorithmic problems and multi-step reasoning.

Installation

/plugin marketplace add qduc/advisor-skills

Skills

Consult

Get second opinions from other models and agents:

  • Strategist Agent: Outside advice on strategic decisions and architectural trade-offs
  • Codex: Reasoning through hard logic problems
  • Gemini: Online research for current information, documentation, and real-world examples

Use when stuck on difficult problems, validating approaches, or needing fresh perspectives.

Installation:

/skill install consult

Strategist Cooperation

Guidelines for effective collaboration with the Strategist agent, including patterns for session continuity, preparing context, and handling advice.

Installation:

/skill install strategist-cooperation

Agents

Strategist

An agent that looks at a problem from outside, without seeing the existing code, implementation details, or project constraints.

Key capabilities:

  • Judges options on merit, ignoring technical debt and sunk costs
  • Reasons from first principles
  • Plays devil’s advocate: challenges assumptions and proposes alternatives

Use when:

  • Stuck on architectural decisions and going in circles
  • Team debates with strong conflicting opinions
  • Evaluating major refactoring or rewrite decisions

It points out hidden assumptions and gives you a way to decide, instead of a prescribed answer.

Why This Approach?

  • Cost: A lighter model does the main work; expensive models only see the critical decisions.
  • Fewer hallucinations: Facts get cross-checked against more than one source.
  • More perspectives: You hear from several models before committing to a big change.
  • Right model for the job: Each task goes to the model that’s good at it.

Docs and contribution guidelines are in the GitHub repository.