Multi-Model Review of specs, plans and tasks for clarify and analyze commands

[jaycangel]

First off, I want to say how much I appreciate the addition of the clarify and analyze commands - they’re a major step forward for spec-kit and have already improved the quality of my planning and outputs.

What I’ve started doing manually is asking other LLMs (Gemini, Codex, GitHub Copilot, etc.) to review, critique, and suggest simplifications or mitigations for the implementation plans that Claude (my primary llm) generates. The process looks roughly like this:

1. Ask Gemini and Codex to review the generated outputs and produce review files like @review-{yyyymmddhhss}-{model}.md.
2. Ask Gemini to then review those independently generated reviews, cross-reference them, and verify the specs, tasks, and plans.
3. i do this by asking Claude to shell out and run the cli's directly.

This approach has hugely improved the quality and robustness of my plans.

Why this matters

Each base LLM has its own strengths and weaknesses. By combining them together in a structured way, I’ve been able to get more reliable, nuanced, and practical plans than by relying on a single model. In my case:

• Gemini CLI provides a generous free allowance for Gmail users.
• Codex CLI offers good paid capacity at ~$20/month.

It feels like a waste not to leverage this compute that’s already available to many users.

Proposed Feature

It would be amazing if clarify and analyze could support a more systematic way of incorporating multi-model reviews — either natively or via configuration. For example, spec-kit could:

• Automatically invoke additional review agents (Gemini, Codex, etc.) on generated outputs.
• Save their responses in timestamped review files.
• Run a cross-review step to compare and validate plans across models to find major improvements, gotchas and missing details.

This would give users the benefit of multi-model collaboration in a clean, integrated workflow, while still preserving spec-kit’s core strengths.