Refactoring Safety Planner

The problem

Large-scale refactors often fail because developers underestimate coupling or break undocumented behavior. Manual planning leads to "big bang" changes that are difficult to review and impossible to roll back safely.

What it does

• Analyzes current source code and dependency graphs to identify structural risks.
• Maps stable behavior contracts including APIs, side effects, and data persistence.
• Generates a staged, multi-step implementation plan using patterns like Branch by Abstraction or the Strangler Fig.
• Defines explicit verification gates and rollback triggers for every stage of the change.
• Separates architectural movement from cleanup to ensure behavior remains identical during migration.

Why this beats prompting it yourself

Generic LLM prompts often suggest destructive rewrites or ignore operational risks like side-effect duplication. This skill enforces a strict safety rubric that prioritizes reversibility and data integrity over aesthetic code changes.

Use cases

• Splitting a bloated monolithic service into modular packages.
• Replacing a core library or database driver without breaking consumers.
• Modernizing legacy architecture while maintaining feature parity.
• Extracting logic into a shared module across multiple repositories.

Known limitations

Does not execute code changes or run tests directly. Requires existing source code or architecture documentation as input to produce an accurate plan.