adr-review

Coordinate six specialized agents through structured ADR review debate rounds.

41|13|Updated Dec 14, 2025

Search Tags:#architecture #multi-agent #orchestration #consensus #ADR #MADR

Ready to run

npx skills add https://github.com/rjmurillo/ai-agents --skill adr-review

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: adr-review
Source: https://github.com/rjmurillo/ai-agents/tree/main/.claude/skills/adr-review
Command: npx skills add https://github.com/rjmurillo/ai-agents --skill adr-review

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Multi-agent debate orchestration for Architecture Decision Records. This skill coordinates six specialized agents through structured review rounds until consensus (or up to 10 rounds).

Core Features & Use Cases

Role-based agent orchestration: architect, critic, independent-thinker, security, analyst, high-level-advisor
Phase-based process: Independent Review, Consolidation, Resolution, Convergence Check
Conflict resolution and scope split guidance to maintain clarity and alignment
Debate artifacts saved to .agents/architecture/ with a formal ADR lifecycle

Quick Start

Trigger adr-review for .claude/skills/adr-review/SKILL.md to start a multi-agent ADR cycle.

Frequently Asked Questions about adr-review

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

FAQPage Schema

How do multi-agent ADR reviews work for architecture decisions?▼

Multi-agent ADR review coordinates six specialized agents—architect, critic, independent-thinker, security, analyst, and high-level-advisor—through structured debate rounds until consensus. The process runs through Independent Review, Consolidation, Resolution, and Convergence Check phases, with debate artifacts saved for traceability.

When should I use orchestrated consensus for Architecture Decision Records?▼

Use orchestrated multi-agent ADR review when creating or updating architecture decisions and need conflict resolution, diverse perspectives, and documented reasoning. It's essential when security, technical trade-offs, and organizational alignment matter and formal traceability is required.

How do I trigger a multi-agent ADR review process?▼

Trigger adr-review on your ADR file to start a multi-agent orchestration cycle. The process automatically coordinates agent roles through phase-based rounds, applies consensus criteria, and outputs debate artifacts to .agents/architecture/ for full lifecycle documentation.

What happens if agents disagree during an ADR review?▼

The orchestration applies conflict resolution and scope-split guidance to maintain clarity and alignment. Agents iterate through structured rounds—up to 10 maximum—until consensus emerges or documented disagreement surfaces with rationale captured in artifacts.

Can ADR reviews handle security and analyst perspectives simultaneously?▼

Yes. The multi-agent framework includes dedicated security and analyst agents alongside architect, critic, independent-thinker, and high-level-advisor roles. All perspectives contribute through structured debate phases to ensure comprehensive decision validation.