safe-refactor

Safely orchestrate large-scale code refactors using a two-tier orchestration model that maps scope, builds safety nets, implements stepwise changes, and verifies behavior through gated phases to protect observable behavior.

Preserve observable behavior during refactoring by establishing characterization tests as a safety net before implementation, then using committed-reviewed results and hard exit gates to validate each step.

The two-tier orchestration model uses an orchestrator to map the plan and spawn a dedicated refactor-worker subagent, driving stepwise implementation through a live working-notes plan for traceability and rollback.

Execute automated refactoring with full traceability by maintaining a live working-notes plan that tracks each deterministic step, ensuring rollback capability and committed-reviewed results throughout the workflow.

A risk-aware code refactoring workflow enforces a four-phase process: Map the scope, build the Safety Net, Implement changes stepwise, and Close with verification, using hard exit gates at each phase.

Use a dedicated refactor-worker for code quality improvements when changes touch many files or modules, requiring deterministic step execution and gated verification to maintain safety during large-scale transformations.