My AI Agent Kept Recommending Abandoned Repos. So I Built RepoCrunch.

I've been building AI agents that evaluate GitHub repos — picking dependencies, reviewing codebases, comparing tools. The problem? They hallucinate. Ask an LLM "is this repo well-maintained?" and you'll get a confident answer based on vibes. Not data...