sdd-plan-review

Coordinate multi-model reviews of SDD specifications into categorized Markdown and JSON reports.

9|3|Updated Oct 24, 2025
One-click install
npx skills add https://github.com/tylerburleigh/claude-sdd-toolkit --skill sdd-plan-review
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: sdd-plan-review
Source: https://github.com/tylerburleigh/claude-sdd-toolkit/tree/main/skills/sdd-plan-review
Command: npx skills add https://github.com/tylerburleigh/claude-sdd-toolkit --skill sdd-plan-review

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill prevents costly implementation errors by identifying design flaws, security risks, and feasibility issues in specifications before coding begins. It combats LLM sycophancy by demanding critical analysis from multiple AI perspectives, ensuring genuine and actionable feedback.

Core Features & Use Cases

  • Multi-Model Validation: Leverages Gemini, Codex, or Cursor-Agent CLIs in parallel for diverse architectural, security, and feasibility perspectives.
  • Anti-Sycophancy Prompts: Engineered with prompts that assume problems exist and demand critical analysis, fighting LLM tendencies to agree.
  • Actionable Feedback: Provides clear, prioritized issues (Critical, High, Medium, Low) with suggested fixes and an overall recommendation.
  • Use Case: Before committing to a complex feature, use this skill to get an unbiased, multi-AI review of your JSON specification, ensuring all angles (security, architecture, feasibility) are covered and issues are caught when they're cheapest to fix.

Quick Start

List available AI tools for review

sdd list-review-tools

Perform a full, comprehensive review of your spec

sdd review specs/active/user-auth-001.json --type full

Save the review report to a Markdown file

sdd review specs/active/user-auth-001.json --output review-report.md

Frequently Asked Questions about sdd-plan-review

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

FAQPage Schema
How do I catch design flaws in my specification before implementation starts?

Specification review identifies architectural issues, security risks, and feasibility problems early by coordinating multi-model analysis across your SDD. This approach consolidates diverse AI perspectives into prioritized, categorized feedback (Critical, High, Medium, Low) before coding begins, preventing costly rework.

Can I use multiple AI models to review my JSON specification at once?

Multi-model review leverages parallel analysis from tools like Gemini, Codex, or Cursor-Agent CLIs to provide independent architectural, security, and feasibility perspectives. Parallel review combats agreement bias and delivers structured, actionable feedback consolidated into Markdown and JSON reports.

What does a specification review report include?

Review reports categorize findings into Missing Information, Design Concerns, Risk Flags, and Enhancement Suggestions, each prioritized by severity. Output includes structured Markdown and JSON formats with clear issue descriptions, suggested fixes, and an overall recommendation to guide downstream implementation.

How do I prevent AI reviewers from just agreeing with my spec?

Anti-sycophancy prompting assumes problems exist and demands critical analysis, ensuring genuine evaluation rather than surface-level approval. This engineering counters LLM tendencies to defer and produces unbiased, multi-perspective feedback grounded in architectural, security, and feasibility rigor.

Does specification review work for complex features or only simple specs?

Specification review handles complex, multi-faceted features by decomposing analysis across architecture, feasibility, risk, and verification dimensions. It's designed for draft SDDs before implementation, particularly useful when feature scope, security requirements, or integration points warrant pre-coding validation.