Assess, compare, strengthen, and synthesize outputs from coding AI agents. Use when Codex needs to evaluate multiple agent plans, task lists, implementation approaches, code review notes, diffs, or other engineering outputs from the same prompt; identify weaker reasoning, missing assumptions, obsolete or risky technology choices, logic gaps, test gaps, and unsupported claims; transfer useful context from stronger outputs to weaker agents; or merge several outputs into a better final plan or task breakdown.
