Self-Review Quick

Run a quick self-review for AI experiment designs by submitting the design as a JSON string with the artifactType set to experiment-design. This triggers a single LLM call that unites Critic, Defender, and Judge roles to output a verdict and reasoning.

Multiagent self-review for artifact validation is a process that unites Critic, Defender, and Judge agents into a single LLM call. It evaluates gaps, ideas, or experiment results to output a verdict and reasoning, reducing loop time compared to slow multi-step reviews.

Yes, you can validate experiment results using a single LLM call by providing the results as a JSON string with artifactType set to experiment-result. The unified Critic-Defender-Judge mechanism will output a SelfReviewResult containing a verdict and reasoning.

Single-call self-review operates without external dependencies, running the Critic, Defender, and Judge evaluation internally within one LLM call. It accepts a JSON artifact string and outputs a verdict, making it lightweight for iterative validation.

Use a one-shot self-review during iterative AI development workflows when you need lightweight, quick quality checks. It prevents slow multi-step reviews by uniting Critic, Defender, and Judge into a single pass, rapidly validating gaps, ideas, or experiment artifacts.

LLM quality assurance validation supports artifact types including gap, idea, experiment-design, and experiment-result. You must provide the artifact as a JSON string alongside one of these artifactType values to trigger the self-review process.