review-plan

name	review-plan
description	Multi-model peer review — sends spec + plan to external AI models for independent critique before implementation.
argument-hint	Optionally pass a specific feature dir (e.g., specs/003-auth-flow)
compatibility	Requires spec-kit project structure, .env with API keys (OPENAI_API_KEY, GEMINI_API_KEY)
metadata	{"author":"jabelk","source":"claude-code-template"}
user-invocable	true

Review Plan — Multi-Model Peer Review

Send the current feature's spec + plan to external AI models (OpenAI gpt-5.3-codex, Gemini 2.5 Pro) for independent peer review before implementation begins.

Data Sensitivity Warning

Before running, consider whether the spec/plan contains sensitive data (PII, credentials, proprietary business logic). The review sends content to external API providers. If the project handles sensitive data, confirm with the user before proceeding.

When to use

Run this AFTER /speckit-plan and /speckit-tasks complete, BEFORE /speckit-implement. This is the gate between planning and coding.

How it works

1. Run ./scripts/review-plan.sh $ARGUMENTS from the repo root
2. The script auto-detects the most recently modified specs/*/plan.md if no argument is given, or pass a specific feature dir (e.g., specs/003-auth-flow)
3. It sends the spec, plan, tasks, data model, and constitution to both models in parallel
4. Each model rates the plan GREEN/YELLOW/RED and lists top issues

After the review

Present both reviews to the user with a summary:

• If all GREEN: "All reviewers approve. Safe to implement."
• If any YELLOW: "Issues flagged — review the concerns below and decide whether to fix before implementing."
• If any RED: "At least one reviewer recommends rethinking the approach. Address the RED concerns before proceeding."

If the user wants to proceed despite warnings, that's their call. The review is advisory, not blocking.