Code Reviewer v1.0.2

A collection of agents for performing comprehensive, evidence-based code quality reviews across multiple programming languages and frameworks.

Prerequisites

• VS Code with the GitHub Copilot Chat extension installed and active.

Installation

Install via the VS Code Chat Plugin Marketplace using the dimpletz/prompts-collection marketplace source and enable the code-reviewer plugin.

Usage

Open Copilot Chat, select one of the code reviewer agents, and provide the diff file path or code to review.

# Code Reviewer (diff-based, full pipeline)
"Review the diff at /tmp/diff-files/my-feature.diff"
"Run a code review on C:\work\changes.diff"

# General code review agents (inline code, files, folders)
"Review this PHP class for security vulnerabilities and PSR-12 compliance"
"Analyze the src/services folder for performance bottlenecks and SOLID violations"
"Check this TypeScript component for best practices and accessibility issues"

Components

Code Reviewer

Orchestrates the full diff-based code review pipeline. Accepts a diff file, splits it into reviewable chunks using the diff-chunker skill, detects the project-level primary language, and dispatches the Language Rules Auditor sub-agent once per chunk. The Language Rules Auditor uses the code-review-report-appender skill to append its findings directly to the shared report file. The report is saved to CODE_REVIEW_REPORT_DIR when set, otherwise to <workspace root>/code-reviews/. Report filename format: <workspace-dir-name>-<diff-basename>-<YYYY-MM-DD-HH-mm>.md.

Report structure:

• Header with reviewer name and date
• [TOC] placeholder
• Ignored files section (if any)
• Unreviewed oversized files section (if any)
• Per-file review sections appended by the Language Rules Auditor via the code-review-report-appender skill

Language Rules Auditor

Sub-agent dispatched by the Code Reviewer orchestrator for each diff chunk. Detects which languages appear in the chunk, invokes the review-rules-provider skill to load the applicable rule set (cross-cutting + project primary-language + chunk-detected language rules), and applies those rules to every changed file. For files whose language has no matching rule file, the auditor applies cross-cutting rules and falls back to built-in expertise. Uses the code-review-report-appender skill to append findings directly to the shared report file.

Code Quality Reviewer

Expert code quality reviewer with deep knowledge of PHP (Magento 2, Laravel, Symfony), .NET (C#, ASP.NET Core), JavaScript/TypeScript (React, Vue.js, Angular, Node.js), Python (Django, Flask, FastAPI), CSS/SCSS, SQL, and more. Reviews individual files, selections, modules, or entire codebases against coding standards, SOLID principles, security best practices, and performance guidelines. Produces structured, prioritized findings grouped by category.

Comprehensive Code Quality Reviewer

Performs thorough, evidence-based code quality analysis against established quality standards and all non-functional requirements (NFRs): security, reliability, performance, scalability, maintainability, usability, portability, compatibility, compliance, and localization. Every finding is backed by a verified URL to official documentation. Generates a Confluence-ready Markdown report saved to the workspace root with a timestamp in the filename. Includes an executive summary, summary statistics table, production readiness verdict, and detailed findings with code snippets.

Key guardrails:

• Only reviews code — will not write new features or fix code directly.
• Asks for clarification before proceeding if the language or scope is ambiguous.
• Skips or marks findings as unverified if official documentation cannot be found.

Universal Code Quality Reviewer

Language-agnostic reviewer that applies a consistent four-step process (Identification → Analysis → Prioritization → Reporting) across PHP, JavaScript/TypeScript, Python, C#/.NET, Java, Go, Ruby, and SQL. Findings are categorized as Critical, High, Medium, or Low and include the exact file location, code snippet, explanation of the problem, and a recommended fix with a code example. Strictly avoids hallucinating URLs or references.

Diff Chunker

Splits a large diff file into smaller chunk files for code review by running a companion script. Diff sections whose line count exceeds CODE_REVIEW_DIFF_CHUNK_MAX_LINES are saved as standalone oversized files. All output files are placed in the same directory as the input diff file, named <original>-chunk-<n>.diff for chunks and <original>-<sanitized-section-path>.diff for oversized sections.

Review Rules Provider

Loads and concatenates the applicable rule set for a review pass. Always includes the cross-cutting (language-agnostic) rules and the project primary-language rules. Appends rules for any additional languages detected in the diff chunk. Deduplicates rule files so each is included at most once. Invoked by the Language Rules Auditor; not intended for direct use.

Code Review Report Appender

Appends markdown content to an existing code review report file using a platform-appropriate terminal command. Invoked by the Language Rules Auditor at the end of each chunk review to write its findings to the shared report. Requires the report file to exist — the Code Reviewer orchestrator creates it before dispatching any sub-agents. Not intended for direct invocation.

Hook

Author

Ron Webb