working-with-legacy-code

Working with Legacy Code is a software-craft reference skill that teaches how to safely change and test untested codebases, distilled from Michael C. Feathers' book 'Working Effectively with Legacy Code'. It reframes legacy code as simply code without tests and addresses the core dilemma that you need tests to change code safely but must often change code to get tests in place. The remedy it teaches is the cover-and-modify discipline: get code into a test harness, pin current behavior with characterization tests, then change one verifiable step at a time instead of a rewrite.

The skill provides a change algorithm (identify change points, find test points, break dependencies, write tests, then change and refactor), a scoring rubric from 0 to 10 for rating how safely changes are made, and a framework covering seams, characterization/golden-master tests, sprout and wrap methods, and dependency-breaking techniques for constructors, statics, and singletons. Reference files add detailed case studies (for example getting an 800-line untested service class or a singleton-ridden Python pricing module under test), plus focused guides on characterization tests, the change algorithm, dependency breaking, and seams. It explicitly separates behavior changes from refactoring so failures stay diagnosable.

Target users are developers who have inherited messy or untested code, are afraid to modify a tangled module, or need to add features or fix bugs while building incremental test coverage. It points to companion skills for refactoring already-tested code and for day-to-day code quality.