What problem does it solve? Reviewing pull requests manually is time-consuming and inconsistent, and low-signal feedback often buries the issues that actually matter. This Skill produces focused, high-confidence review findings on GitHub pull requests so reviewers can act on real bugs instead of noise. ## Core Features & Use Cases - Diff-Based Review: Reads PR metadata, changed files, and diffs through the gh CLI and local git, reviewing only changed behavior. - Confidence-Scored Findings: Scores each candidate finding from 0 to 100 and keeps only those at or above a configurable threshold (default 80), filtering out speculative noise. - Structured Output: Reports findings with severity, confidence, file and line references, impact rationale, and the smallest reasonable fix, followed by test gaps and a summary. - Use Case: Before merging a feature branch, ask for a review of the open PR and receive a concise list of correctness bugs, security regressions, and missing tests suitable for posting as a PR comment. ## Quick Start Review the open pull request in this repository and list any high-confidence findings with file and line references.
