---

name: architecture-review description: Run an iterative architecture review loop with critique and revision

Architecture Review Skill

Iteratively critiques and improves an architecture document through multiple rounds of review.

When to Use

Use this skill when you want to:

• Review and improve an architecture document
• Get structured critique of a design
• Iteratively refine a proposal
• Catch issues before implementation

Input Format

CopyDocument: [Your architecture document or design]
Max Iterations: [Optional - max review rounds, default 3]

Example

CopyDocument:
# User Authentication Service

## Overview
JWT-based auth with refresh tokens...

## Components
- Auth API (Express.js)
- Redis for session cache
- PostgreSQL for user data

## Flows
1. Login: email/password → validate → issue JWT
2. Refresh: refresh_token → validate → new JWT
...

Max Iterations: 3

What Happens

1. Document is sent to the LangGraph design-review endpoint
2. AI generates critique identifying issues and suggestions
3. If issues found, document is revised
4. Critique runs again on revised version
5. Loop continues until approved or max iterations reached

Response Format

Copy## Final Document
[The improved document after all revisions]

## Review Summary
- Iterations: 3
- Final Status: Approved

## Critique History
### Iteration 1
- Issues: Missing rate limiting, no token revocation
- Suggestions: Add rate limit per IP, implement token blacklist

### Iteration 2
- Issues: Blacklist approach won't scale
- Suggestions: Use short-lived tokens + sliding refresh

### Iteration 3
- Status: Approved
- Minor suggestions: Consider adding...

Implementation

Call the LangGraph API at: http://localhost:8100/graphs/design-review

Copycurl -X POST http://localhost:8100/graphs/design-review \
  -H "Content-Type: application/json" \
  -d '{
    "document": "# User Authentication Service\n\n...",
    "max_iterations": 3
  }'