review-seq

What problem does it solve?

Provides a resilient, disk-backed code review orchestration that prevents findings from being lost to context overflow by persisting per-tool outputs, compacting context between waves, and supporting resume after interruptions. It ensures every automated analysis gets full attention and that large or long-running reviews remain deterministic and recoverable.

Core Features & Use Cases

• Batched analysis with persistence: Run analysis tools in parallel batches or sequentially while writing findings to a claude_session directory so no results depend on ephemeral context.
• Resume and compaction points: Recover interrupted runs and compact context between batches to avoid context-window loss while preserving a full audit trail.
• Scoring, root-cause, and fix plan generation: Aggregate findings, deduplicate and score them, run root-cause analysis waves, and produce a prioritized fix plan with anchors for safe automated fixes in review-fix mode.
• Practical scenarios: Use for pre-merge gate checks, automated review-and-fix workflows, or multi-session reviews on large Python changesets where tools like ruff, ty, semgrep, and pytest are used.

Quick Start

Run the review-seq pipeline on the current branch in review-fix mode to persist findings, score and analyze root causes, and produce a prioritized fix plan.