AI Code Review Suite
Automated code review for Bitbucket PRs across all domains
The Problem
Manual Code Review Bottleneck
Code reviews pile up in the queue because reviewers are busy with their own work. PRs sit for hours or days waiting for human attention, blocking merges and slowing the entire team's velocity.
Domain-Blind Reviews
Generic code review checklists miss domain-specific concerns. A backend API review needs different checks (SQL injection, N+1 queries) than a PLC review (sequence deadlocks, CODESYS architecture). One-size-fits-all reviews catch syntax but miss substance.
No Structured Feedback
Review comments are scattered across PR threads with no consistent format. Critical issues blend in with nitpicks, making it hard to prioritize fixes. There's no way to track review quality or patterns across the team.
PRs waiting hours/days for human reviewer availability
Automated review posted within minutes of PR creation
Same generic checklist applied to backend, frontend, and PLC code
Domain-specific review lenses auto-selected by file extension analysis
Unstructured comments mixed across PR threads
Structured Markdown report with severity levels posted directly to PR
Approach
Domain-Aware AI Review Engine
The system analyzes Bitbucket PR diffs and auto-detects the review domain from file extensions — .py/.sql files trigger backend review (API design, database patterns, security), .tsx/.css files trigger frontend review (component architecture, state management, accessibility), and .st/.pou files trigger PLC review (CODESYS structure, sequence deadlock detection). Each domain has its own review template with domain-specific checklist items and severity levels. The AI model receives the diff along with the domain-specific prompt, producing a structured Markdown review that gets posted as a PR comment via the Bitbucket API. The skill operates standalone — it works in any project with just environment variables configured, requiring no project-specific setup.
BDD Pipeline Flow
Key Features
Auto-Detect Domain
Analyzes file extensions in the PR diff to automatically determine the review domain. Backend (.py, .sql, .go), Frontend (.tsx, .css, .html), and PLC (.st, .pou) each activate specialized review templates with domain-specific checklist items and focus areas.
Structured Review Template
Reviews follow a consistent Markdown template with sections for summary, critical issues, suggestions, and nitpicks. Each finding includes severity level (critical/warning/info), file location, and actionable fix suggestion, making it easy to prioritize and resolve feedback.
Bitbucket Integration
Fetches PR details and diffs via the Bitbucket REST API, then posts the structured review as a PR comment. Supports both Bitbucket Cloud and Server. The review appears inline with the PR conversation, requiring no context switch for developers.
Multi-Language Support
Handles code in Python, TypeScript, Go, SQL, Structured Text, and more. The AI model adapts its review focus based on the language — checking for type safety in TypeScript, memory management in Go, and timing hazards in PLC code.
Architecture
Results
Backend, Frontend, and PLC specialized reviews
File extension analysis determines review type
Python, TypeScript, Go, SQL, ST, and more
Standalone skill — env vars only, no project config