migrate-research

To improve migration converter fidelity, this Skill iteratively updates Rust code and validates changes against a priority style corpus. It runs an evidence-driven loop that clusters mismatches, applies targeted fixes, and uses cargo nextest to retain improvements or revert failures.

Grouping and classifying migration mismatch root causes is handled by clustering observed failures into distinct groups. The Skill classifies whether root causes originate from the converter, style, processor, or adjudication layers, enabling targeted code fixes based on the identified pattern.

Yes, the fidelity testing loop automatically reverts failed converter code changes. It uses cargo nextest validation to evaluate each change attempt, permanently keeping improvements that pass while automatically reverting any modifications that fail the regression tests.

Validating converter code changes against a style corpus requires running the full evidence loop. The Skill measures fidelity outcomes across the priority style corpus, uses cargo nextest for regression testing, and generates a decision-quality report to confirm if the changes should be kept or reverted.

Yes, you can resume an interrupted automated regression testing session for migration styles. The Skill supports repeated migration mismatch clustering and resume sessions, allowing you to continue the converter improvement loop across multiple styles until a stop or escalation verdict is reached.

Migration output has formatting and data-mapping differences because the converter logic deviates from high-fidelity reference outputs. This Skill reduces that gap by applying an autonomous evidence-driven loop of converter code changes and validation against a priority style corpus to iteratively raise fidelity.