AI Refactoring Agent
CodeHuskAI
Refactoring that goes through review.
An agent that reads a repository before it changes one, then proposes patches with tradeoffs an engineer can review.
- AIANALYSIS
- GitCONTEXT
- CIFEEDBACK
Testing whether an agent can propose a refactor an engineer actually trusts.
Case study
Context
Reads before it edits.
The agent maps repository structure, local conventions, and likely breakage before proposing a patch.
Review
The engineer keeps control.
It explains the tradeoff behind a refactor instead of treating the patch as self-evident.
Feedback
CI is part of the loop.
Git and CI context are part of the feedback surface, so review has concrete signals to inspect.