multi-model-code-review
Multi-Model Code Review
Run code changes through multiple AI models to catch issues before human review. Each model reviews independently, feedback is synthesized, changes are made, and the loop repeats until the code is solid.
Prerequisites
At least one agentic CLI must be installed and authenticated:
| CLI | Verify | Print-mode flag |
|---|---|---|
| Cursor Agent | agent --version |
agent --model <model> --print "<prompt>" |
| Claude Code | claude --version |
claude --model <model> -p "<prompt>" |
| OpenAI Codex | codex --version |
codex --model <model> -q "<prompt>" |
If cursor's agent is installed, always prefer using that and specifying models rather than mixing and matching CLI utilities.
Workflow
More from cosmastech/skills
multi-model-planning
>-
17php-conventions
Personal PHP conventions enforced when creating or modifying PHP files. Covers strict types, function imports, testing philosophy, class design, and observability. Activate whenever working on PHP code.
15laravel-conventions
Personal Laravel conventions enforced when creating, modifying, or planning code that will touch code utilizing the Laravel framework.
13planning-conventions
Personal conventions for planning software changes. Covers planning doc structure, failure scenario analysis, refactoring sequencing, observability planning, and multi-model review. Activate when planning features, writing design docs, or scoping technical work.
3