What problem does it solve?
This Skill automates the tedious, error-prone process of reviewing branch changes and applying concrete fixes for critical and warning issues so developers can ship safer code faster. It centralizes multiple analysis tools and reviewer personas into a repeatable pipeline that reduces false positives via semantic voting and enforces style and safety rules.
Core Features & Use Cases
- Scope discovery: Collects changed, staged, and uncommitted Python files from the current branch and skips work when no targets exist.
- Parallel analysis: Launches seven agents concurrently (static analysis, web confirmation, three semantic passes with voting, grumpy review, and style checks) and consolidates canonical findings.
- Semantic voting & consolidation: Intersects multi-pass semantic reviewers to reduce false positives and deduplicates findings by file:line:source before building a prioritized fix list.
- Safe automated fixes: Applies only listed, unambiguous edits via a dedicated fixer agent, marks ambiguous cases MANUAL, and avoids editing files outside the branch.
- Test validation and reporting: Runs unit tests and diff-coverage, then produces a two-column report of applied fixes and manual items.
- Disk-persisted mode: Supports batch or sequential disk-backed pipelines for large change sets to avoid context overflow.
Quick Start
Run the review-fix skill to analyze changed Python files in the current branch and apply safe automated fixes.