GPT-4 Code Review Agent

A GitHub Action that uses GPT-4 to automatically review every pull request in your repository — identifying security vulnerabilities, code quality issues, performance bottlenecks, and style violations, then posting detailed review comments directly on the PR.

What's Included — Full Source Code:

• Complete TypeScript GitHub Action
• GPT-4 review engine with specialised prompts
• Security vulnerability scanner
• Self-hosted runner option
• Full documentation

Core Features:

• Automatic PR reviews — triggers on every pull request, analyses the diff, and posts line-by-line review comments
• Security scanning — detects OWASP Top 10 vulnerabilities, hardcoded secrets, SQL injection, XSS, and insecure dependencies
• Code quality — identifies code smells, complexity issues, dead code, and missing error handling
• Performance analysis — flags N+1 queries, memory leaks, unnecessary re-renders, and unoptimised algorithms
• Style consistency — checks naming conventions, file organisation, and project-specific patterns
• Suggested fixes — provides concrete code suggestions, not just complaints

Configuration:

• Review strictness level (lenient, standard, strict)
• Focus areas (security-only, quality-only, or all)
• File pattern filters (only review src/, ignore tests/)
• Custom rules via natural language ("always check for SQL injection in database files")
• Language support: TypeScript, JavaScript, Python, Go, Rust, Java, C#

Output:

• Line-by-line GitHub review comments
• PR summary comment with severity breakdown
• Approve / Request Changes based on findings
• JSON webhook for CI/CD integration
• Slack/Discord notification with review summary

Tech Stack:

• TypeScript GitHub Action
• OpenAI GPT-4 / GPT-4o API
• GitHub REST + GraphQL API
• Tree-sitter for AST parsing
• Docker for self-hosted runner

Ship safer code. Catch bugs before humans review. Full source code.