auto-review-loop

Automated iterative research review uses a secondary Codex agent to assess content, apply fixes, and re-review until a positive assessment or maximum rounds are reached. This loop handles review, implementation, and documentation autonomously.

Yes, an autonomous research workflow can include an optional human checkpoint feature. This allows you to provide input or approve actions at each review round before the Codex agent continues the iteration.

State persistence saves the review loop's progress to recover from context compaction. This ensures the automated research workflow resumes exactly where it left off without losing assessment history or implementation state.

Running an automated review loop requires a secondary Codex agent and access to your research context. No additional dependencies are needed, as the skill operates independently using its internal scripts.

The best way to document an autonomous review process is to let the loop automatically record each round. It documents the assessment, actions taken, and results, ensuring a complete history of the iterative improvement.