aayush2724/LeadForge ? reverse-engineered prompt

Build me a reproducible Python pipeline that takes raw lead exports, cleans and merges them, filters out bad fits, enriches the remaining companies, scores each lead, and generates personalized outreach for the best ones.

I want it to work end to end from one command, with CSV outputs at each stage so I can inspect the results and rerun it monthly with fresh data. It should handle sources like Apollo, LinkedIn, seed lists, and other company research inputs, then dedupe by domain, remove obvious competitors or disqualified companies, and produce a final lead list split into hot, warm, and cold tiers based on a clear scoring system.

Please also generate ready to send email and LinkedIn message drafts for the top leads, plus a simple A/B variant set for testing. If it makes sense, add a lightweight workflow or automation entry point so the whole thing can be triggered easily. Use current docs online if you need to.