Can it automatically update my code review skills?

The skill currently generates detailed recommendations and specific code snippets that you can apply to your skill files, providing a safe, human-in-the-loop improvement process.

What is the primary benefit of the Review Skill Improver?

It reduces developer fatigue by identifying and fixing automated rules that frequently flag valid code as errors, ensuring your AI code reviews remain helpful and high-signal.

Does this support specific frameworks like Pydantic AI or React?

Yes, it is designed to analyze patterns across any framework. It can identify when framework-specific patterns are being incorrectly flagged and suggest updates to the relevant skill.

How does it determine which rules need to be changed?

It flags rules with a rejection rate higher than 30% and performs pattern analysis on the rationales provided by developers to suggest specific exceptions or logic adjustments.

What kind of input does this skill require to function?

It requires structured feedback logs in an enhanced schema format, which track whether review suggestions were accepted or rejected by the user and the reasoning behind those decisions.

Review Skill Improver

Name: Review Skill Improver
Author: existential-birds

byexistential-birds

•

コラボレーションツール

Optimizes code review accuracy by analyzing feedback logs to identify patterns, reduce false positives, and refine automated review rules.

概要

The Review Skill Improver is a specialized tool designed to create a continuous improvement loop for automated code reviews within Claude Code. By processing structured feedback data, it identifies rules with high rejection rates, detects missing coverage, and provides actionable recommendations to update skill definitions. It is particularly useful for development teams that want to move beyond generic linting toward high-precision, context-aware code analysis based on empirical performance data and real-world developer feedback.

主な機能

Aggregates feedback by rule source to calculate rejection and acceptance rates
Differentiates between intentional design decisions and actual coding errors
14 GitHub stars
Generates specific markdown recommendations for skill modifications and fixes
Identifies high-rejection rules and common false positive themes automatically
Produces structured improvement reports with data-backed evidence and impact projections

ユースケース

Tailoring code review skills to match specific project-level framework patterns
Reducing noise in automated PR reviews by refining over-active or outdated rules
Creating a data-driven documentation cycle for internal coding standards and best practices

概要

主な機能

Aggregates feedback by rule source to calculate rejection and acceptance rates
Differentiates between intentional design decisions and actual coding errors
14 GitHub stars
Generates specific markdown recommendations for skill modifications and fixes
Identifies high-rejection rules and common false positive themes automatically
Produces structured improvement reports with data-backed evidence and impact projections

ユースケース

Tailoring code review skills to match specific project-level framework patterns
Reducing noise in automated PR reviews by refining over-active or outdated rules
Creating a data-driven documentation cycle for internal coding standards and best practices