End-to-end project, artifact, and code review for assessing completion, requirements coverage, architecture, implementation quality, testing, risks, documentation, readiness, and improvement opportunities. Use when Codex needs to audit a software or technical project, review a diff or PR, judge whether work is…
Use when the user asks Codex App to run a project for a duration, keep working automatically, auto-improve a project, continue in the background, use human checkpoints only for risky actions, resume/pause/cancel/status an automation, or run a long autonomous project loop with self-review and reviewer subagents.