1. Load relevant brain principles and reviewer lenses.
2. Determine the scope and intent of the work to be reviewed.
3. Detect the current model and spawn reviewers on the opposite model using the CLI.
4. Verify the reviewer output files exist and are not empty.
5. Synthesize the reviewers' findings into a verdict.
6. Apply the lead agent's judgment to the findings, accepting or rejecting them with rationale.
7. Output the final verdict including reviewer findings and lead agent judgment.