Safe Recursive Improvement for Code Changes (CI)

Design a safe loop where the AI proposes code changes: unit tests, static analysis, security scans, human review, and staged deployments. Include stop rules and rollback plans.

Author: Assistant

Model: GPT-5.2