pair-consult

Coordinate a bounded five-round consultation between Claude Code and Codex.

8|1|Updated Feb 22, 2026
One-click install
npx skills add https://github.com/ccomkhj/skills --skill pair-consult
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: pair-consult
Source: https://github.com/ccomkhj/skills/tree/main/pair-consult
Command: npx skills add https://github.com/ccomkhj/skills --skill pair-consult

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

pair-consult solves the problem of making better decisions on a plan or small coding task by forcing a structured, bounded back-and-forth between two agents until you reach a clear outcome.

Core Features & Use Cases

  • Bounded 5-round consultation: runs a fixed proposal → review → response → re-review → synthesis loop with explicit round artifacts in .consult/.
  • Critique-driven convergence: reviewers provide numbered critiques and the proposer must address every item (including “object” and “partial” verdicts) without skipping.
  • Coding-friendly guardrails: for coding tasks, the reviewer verifies by running tests and inspecting diffs, while the proposer re-runs tests after changes.
  • User-facing final close: produces an R5 synthesis that summarizes what was landed, unresolved tensions, and what you need to decide.

Quick Start

Run pair-consult on your question with: /pair-consult "<your question>".

Frequently Asked Questions about pair-consult

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

FAQPage Schema
How do I get a second opinion on a coding decision without a complex workflow?

A bounded 5-round consultation coordinates a back-and-forth between two agents to resolve a single coding question. It uses a structured proposal, review, response, re-review, and synthesis loop to reach a clear outcome without heavy ceremony.

How does structured pair programming handle disagreements between agents?

Structured pair programming uses critique-driven convergence, where reviewers provide numbered critiques and the proposer must address every item, including partial or objected verdicts, ensuring no feedback is skipped before reaching a final synthesis.

Can I use agent collaboration to verify code changes during a review?

Yes, agent collaboration includes coding-friendly guardrails where the reviewer verifies changes by running tests and inspecting diffs, and the proposer re-runs tests after making modifications to ensure correctness.

What is the best way to manage state during a multi-round agent consultation?

Multi-round agent consultation manages state using explicit round artifacts stored in files like QUESTION.md, STATE.md, and R1.md through R5.md, orchestrated via a handoff script to enforce one narrow action per round.

When should I avoid a bounded 5-round consultation for decision support?

You should avoid a bounded 5-round consultation when your task requires full multi-step planning or involves multiple complex questions, as this workflow is designed to resolve a single user question or small coding task decisively.

Does pair-consult require any external dependencies to run?

Pair-consult requires no external dependencies, but it does need round-state files under a .consult/ directory and a handoff orchestration script to manage the agent wait and transition states.