Automated PR analysis focused on reliability, edge cases, and risky logic paths.
Traditional linters and formatters are great at catching syntax and style issues, but many production issues come from logic errors, risky edge cases, or unexpected execution paths.
PullShark is an experimental automated code review system that analyzes pull requests and highlights potentially risky logic changes before they get merged.
The goal isn’t to replace human review — it’s to help developers catch issues earlier and focus attention on high-risk areas.
- Detecting potentially risky control-flow patterns
- Highlighting edge-case heavy code paths
- Surfacing logic that may lead to resource exhaustion or unstable execution
- Generating review notes and testing suggestions for suspicious changes
- Reducing noisy review comments and focusing on meaningful risks
- GitHub App webhooks
- Background PR analysis pipelines
- LLM-assisted review workflows
- Redis-backed processing
- Async job execution
- Reliability-first backend design
- Node.js
- Docker
- Docker Compose
- GitHub App credentials
- MongoDB instance
- Redis instance
git clone https://github.com/yb175/pullShark.git
cd pullSharknpm installCreate a .env file in the project root and add the required credentials:
MONGODB_URI=
REDIS_HOST=
REDIS_PORT=
GITHUB_APP_ID=
GITHUB_CLIENT_ID=
GITHUB_CLIENT_SECRET=
GITHUB_PRIVATE_KEY=
GITHUB_WEBHOOK_SECRET=
JWT_SECRET_KEY=You can find a more detailed setup guide in the backend documentation.
docker compose up --buildnpm run devThe server should now be running locally.
PullShark is still evolving, but the broader idea is to explore how automated systems can assist engineers in reviewing complex pull requests, identifying risky behavior, and improving developer workflows around code review.
Contributions, feedback, and discussions are always welcome.