drift-prevention

To prevent spec drift in a long implementation session, you can trigger a re-anchoring process that re-reads active specifications, git state, issue status, and review comments to re-establish the ground truth and correct divergent implementation paths.

Session anchoring is the process of periodically re-reading project artifacts like specifications and git state to ensure AI alignment. It is needed when an AI works on a feature for an extended period and risks forgetting original acceptance criteria or introducing scope creep.

You can verify implementation alignment by comparing the current code changes against defined requirements and project artifacts. This drift detection mechanism identifies when the AI's active work diverges from the established acceptance criteria or planned scope.

No dependencies or components are required to use this approach. It functions by directly reading existing project artifacts such as active specifications, git state, issue status, and review comments to re-anchor the current session to the source of truth.

The best way to handle scope creep is to periodically re-anchor the session by comparing current implementation against defined requirements and active specifications. This ensures the AI continuously aligns its code generation with the original project scope.