ce-code-review

Multi-agent code review before a PR is handled by spawning specialized reviewer personas to analyze diffs, merging deduplicated findings into a single actionable report. It coordinates correctness, testing, and maintainability checks to reduce manual review time.

Analyzing branch diffs for correctness and maintainability is done by orchestrating parallel reviewer personas that assess code changes dynamically. The workflow uses a fork-safe base-resolution script to determine the correct scope and merges findings into a single report.

This code review workflow supports standalone checks, branch reviews, and PR targets. It uses a fork-safe base-resolution script to accurately determine the review scope across different checkout states.

You need git, gh, jq, bash, awk, and sed installed to run automated diff analysis. These dependencies support the underlying scripts that resolve base branches, extract diffs, and interface with repository tracking systems.

Multi-agent code review merges findings by deduplicating issues from spawned personas into a structured artifact format. The final report provides a stable action path categorized as safe_auto, gated_auto, manual, or advisory.

Filing residual code review findings as tickets is supported through optional tracker backends using gh. The workflow guides subsequent fixes or ticketing by surfacing residual work identified during the diff analysis.