oracle-review-loop

Automate document review loops by running creator and evaluator sessions in a fixed chat thread using a user-approved 100-point rubric. The deliverable undergoes repeated improvement until it achieves a passing score.

A structured review process for quality assurance evaluates documents against a user-approved 100-point rubric. It uses fixed chat threads to provide feedback and requires repeated improvement until the deliverable passes.

Yes, you need Python scripts for loop management and evaluation. These scripts handle the automated review process, execute the evaluator sessions, and determine if the document meets the rubric criteria.

Yes, the review process utilizes a 100-point rubric that must be approved by the user. This rubric is applied to evaluate deliverables like research plans, READMEs, and design documents.

When a document fails evaluation, the review loop executes repeated improvement cycles. The creator session updates the deliverable based on structured feedback until it passes the rubric criteria.

The review automation relies on fixed chat threads and a 100-point rubric, meaning the process is constrained to the defined rubric criteria. It requires Python scripts and cannot dynamically adjust scoring rules mid-loop.