engineering-quality-critique-panel

Orchestrates independent judge agents that debate and produce a consensus review report.

Updated Jan 28, 2026
One-click install
npx skills add https://github.com/scanady/nexus-skills --skill engineering-quality-critique-panel-scanady
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: engineering-quality-critique-panel
Source: https://github.com/scanady/nexus-skills/tree/main/skills/engineering-quality-critique-panel
Command: npx skills add https://github.com/scanady/nexus-skills --skill engineering-quality-critique-panel-scanady

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? A single review pass often misses failure modes because one perspective cannot cover requirements coverage, architectural soundness, and code quality at once. This Skill runs a multi-agent debate review so completed work gets judged from several independent angles before it ships. ## Core Features & Use Cases - Independent Judge Panel: Spawns 2-4 specialized judge agents (Requirements Validator, Solution Architect, Code Quality Reviewer) that review work in isolation with evidence-based findings. - Debate and Reconciliation: Maps consensus versus conflicts, feeds opposing positions back to judges, and resolves or records genuine disagreements without manufacturing consensus. - Prioritized Consensus Report: Produces one severity-ranked report with a Ship / Fix-first / Rework verdict, judge scorecard, and concrete action items. - Use Case: Before merging a large feature branch, ask for a multi-perspective critique to get independent verdicts on requirements coverage, design soundness, and code quality, with every finding backed by file:line evidence. ## Quick Start Ask the agent to run a critique panel review on your recent changes and return a consensus verdict with severity-ranked findings.

Frequently Asked Questions about engineering-quality-critique-panel

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

FAQPage Schema
How do I get a multi-perspective code review from AI agents?▼

Run a critique panel that spawns independent judge agents, each reviewing the work through one lens such as requirements coverage, architecture, or code quality. Their verdicts are reconciled through debate into a single severity-ranked report with a Ship, Fix-first, or Rework verdict.

How does a multi-agent debate review work?▼

Each judge reviews the work independently without seeing other verdicts, then conflicting positions are fed back to the relevant judges to defend or concede. Resolvable conflicts go to the better-supported position; unresolved ones are recorded as reasonable disagreements.

Can the critique panel fix the issues it finds?▼

No, the panel is strictly report-only and never edits the reviewed work. It recommends concrete action items ranked as Must, Should, and Could, and fixes should be routed to a separate fix or simplify skill afterward.

When should I not use a judge panel review?▼

Skip the panel when work is still mid-flight and changing, when the change is trivial enough for one quick read, or when you want fixes applied directly. Panel overhead only pays off on finished, non-trivial units of work.

Does the critique panel work for non-code artifacts?▼

Yes, the judge lenses can be swapped for documents, plans, designs, or decisions, using perspectives like Audience Fit, Logical Soundness, and Clarity. The panel stays at two to four judges to keep the signal strong.