review-synthesis

To consolidate code reviews, this Skill cross-checks multiple review outputs against the actual codebase. It deduplicates findings and produces a single prioritized report with provenance and branch/PR context.

Blind synthesis is a validation mechanism that operates on anonymized review outputs. It verifies every finding against the actual codebase without source bias or vote-counting to ensure only accurate issues remain.

You can deduplicate code review findings by placing all review outputs in a shared directory. The synthesis pipeline then anonymizes and cross-checks them against the actual code to remove redundant issues.

Yes, you can automate code review validation at scale by using a three-stage pipeline with a blind synthesizer. This approach prevents source bias by verifying anonymized findings against the actual codebase.

To generate a structured classified.json for QA, you need to save anonymous review outputs to a shared directory. The reporter stage of the pipeline processes these to output the final structured file.

You need cross-checking when you have multiple code review outputs and want to ensure every finding is accurate. This validation process prevents false positives by verifying findings against the actual codebase.