research

Generate research documentation and coordinate adversarial plan reviews for GitHub issues.

8|2|Updated Feb 8, 2026
One-click install
npx skills add https://github.com/psaab/xpf --skill research-psaab
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: research
Source: https://github.com/psaab/xpf/tree/main/.claude/skills/research
Command: npx skills add https://github.com/psaab/xpf --skill research-psaab

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This skill prevents the high cost of throwaway code by enforcing a rigorous, multi-perspective architectural review process before any implementation begins.

Core Features & Use Cases

  • Triple-Review Validation: Coordinates adversarial feedback from Claude SMR, Codex, and AGY to stress-test design proposals.
  • Plan-of-Action Generation: Produces a comprehensive, multi-path research document that outlines tradeoffs and recommended implementation strategies.
  • Safety Gate: Ensures that no production code is touched until a plan reaches a converged PLAN-READY state and receives explicit manual approval.

Quick Start

Invoke the research skill by providing the issue number and a brief description to initiate the adversarial review process for your proposed architectural change.

Frequently Asked Questions about research

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

FAQPage Schema
How do I validate architectural soundness before writing any code?

You validate architectural soundness by running an adversarial 3-way plan review process that stress-tests design proposals across multiple perspectives before implementation begins. This prevents the high cost of throwaway code by coordinating feedback to reach a converged, manually approved plan.

What is an adversarial plan review for software architecture?

An adversarial plan review is a structured process that generates detailed research documentation on a dedicated branch and iterates through feedback rounds. It validates design tradeoffs and technical feasibility by coordinating hostile feedback from multiple tools until a converged plan-of-action is reached.

How do I start a 3-way architectural code review for a GitHub tracking issue?

You start a 3-way architectural code review by invoking the process with your GitHub issue number and a brief description. The system then initiates the adversarial review process, generating research documentation on a dedicated branch and iterating through feedback rounds.

Do I need Claude SMR, Codex, and AGY to perform a structured plan review?

Yes, the structured plan review requires integration with Claude SMR, Codex, and AGY to validate design tradeoffs and technical feasibility. These tools coordinate to provide the triple-review validation needed to stress-test your architectural design proposals.

Does the architectural review process modify production source code?

No, the architectural review process does not modify production source code. It operates as a safety gate by generating research documentation on a dedicated branch, ensuring no production code is touched until a plan reaches a converged PLAN-READY state with explicit manual approval.

When should I use a multi-perspective architectural review process?

You should use a multi-perspective architectural review process when proposing complex architectural changes to GitHub tracking issues. It is designed to enforce rigorous validation of design tradeoffs before implementation, preventing the high cost of throwaway code in software engineering projects.