adversarial-multi-model-review

To run an adversarial code review across multiple AI models, initiate the skill with your artifact to start a four-phase process that switches models between phases for independent cross-perspective analysis.

Cross-model adversarial review is a process that evaluates architecture designs by running parallel, cross-checking perspectives across multiple AI models to surface issues a single reviewer might miss.

Yes, you can use this multi-model review process for plans, designs, and architectures. It evaluates engineering, security, operations, and production dimensions to produce a consolidated improvement list.

Model switching works by requiring you to change AI models between Phase 1 and Phase 4, ensuring independent analyses that maximize adversarial value and surface hidden issues.

Intermediate findings are stored outside the repository as artifact-driven outputs, ensuring independent analyses remain uncontaminated before final synthesis produces a consolidated improvement list.

The best way to identify missed risks in software engineering designs is using a structured adversarial review that applies a devil's advocate dimension alongside engineering and security cross-perspectives.