I built an AI agent that scores any GitHub repo 0-100 before you clone it

Developer Sarvar Nadaf has introduced RepoPilot, an AI-powered agent designed to evaluate public GitHub repositories before users commit to cloning them. Built using Amazon Bedrock and AWS Lambda, the tool assigns a 0-100 trust and safety score based on five key dimensions: safety, trust, maintenance, maturity, and documentation. RepoPilot specifically scans for risky installation patterns like 'curl | bash' and identifies fake-star inflation, providing users with a clear risk assessment. Beyond scoring, the agent assists developers by offering plain-language troubleshooting for failed installation commands, aiming to reduce the frustration of first-run errors. The project, which was developed for an AWS Builder Center challenge, is open-source and demonstrates how LLMs can be integrated with deterministic Python tools to provide reliable, fact-based analysis of software projects. RepoPilot is currently available as a live demo, offering a practical solution for vetting code in an era of increasing repository-based security threats.