Working in an unfamiliar repository changes how you define progress. A large idea may be exciting, but a maintainer needs a change that can be understood, tested, and safely owned after you leave.
The most effective patches I have worked on started with an exact failure, not an abstract improvement.
## reduce before you repair
Reproduction is design work. A small fixture or focused test removes assumptions and gives reviewers a shared object to reason about. Once the failure is precise, the implementation usually becomes smaller as well.
This is especially important around CI, infrastructure, and distributed tools where a symptom can be several layers away from its cause.
## review is part of the implementation
A patch is not finished when it passes locally. It is finished when another engineer can see the mechanism, understand the boundaries, and change it later without rediscovering the entire problem.
▊ end of file