spec-review

Dispatch design reviews to Claude, Gemini, and Codex and synthesize feedback.

Updated Feb 5, 2026
One-click install
npx skills add https://github.com/jivebreaddev/skills-claude-agent --skill spec-review-jivebreaddev
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: spec-review
Source: https://github.com/jivebreaddev/skills-claude-agent/tree/main/.claude/skills/spec-review
Command: npx skills add https://github.com/jivebreaddev/skills-claude-agent --skill spec-review-jivebreaddev

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Use the spec-review skill to orchestrate external AI reviewers (Claude, Gemini, Codex) for spec/design reviews, ensuring governance, transparency, and a deterministic workflow that prevents unilateral, biased conclusions.

Core Features & Use Cases

  • Dispatches design review requests to multiple AI reviewers in parallel to obtain independent feedback.
  • Enforces the use of the script-based reviewer invocation to ensure deterministic, auditable reviews.
  • Synthesizes reviewer feedback into a structured advisory with explicit consensus, divergence, concerns, and recommended actions.
  • Supports optional context and decision records to ground reviews in project history and constraints.

Quick Start

Submit a design review request and read the advisory after all reviewers have responded.

Frequently Asked Questions about spec-review

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

FAQPage Schema
How do I orchestrate multi-AI spec reviews to avoid biased design conclusions?

Multi-AI spec reviews prevent biased design conclusions by dispatching requests to Claude, Gemini, and Codex in parallel, waiting for all reviewers to respond, and synthesizing their independent feedback into a single structured advisory.

What is the best way to synthesize feedback from multiple AI reviewers?

The best way to synthesize feedback from multiple AI reviewers is to orchestrate them via a mandatory script, enforce deterministic invocation, and aggregate their independent responses into an advisory detailing consensus, divergence, and recommended actions.

How do I ground AI spec reviews in existing project context and decision records?

You can ground AI spec reviews by providing optional context from .omt/specs/context and related decision records, ensuring the synthesized advisory reflects project history and constraints accurately.

Can I use a script to ensure deterministic and auditable AI design reviews?

Yes, you can ensure deterministic and auditable AI design reviews by enforcing a script-based reviewer invocation workflow, which governs the dispatching process and waits for all AI reviewers before synthesis.

Does multi-AI spec review support parallel dispatching to external models?

Yes, multi-AI spec review supports parallel dispatching to external models like Claude, Gemini, and Codex, allowing them to independently evaluate the design before their feedback is synthesized into a unified advisory.

What is included in the structured advisory returned after an AI spec review?

The structured advisory returned after an AI spec review includes explicit consensus, divergence, concerns, and recommended actions derived from the synthesized independent feedback of all participating AI reviewers.