What is drinks-sommelier?

This project is a skill file that teaches an AI agent to act as a personal beer and wine sommelier. Instead of being a standalone app, it is a set of instructions that plugs into AI coding assistants and agents like Claude Code, Cursor, or OpenCode, giving them a specific job: recommend the best drink for you from whatever options are in front of you. The idea is simple. At a restaurant, pub, or shop, you often face many choices with no clear sense of which one fits your taste. Instead of guessing or searching multiple sites, you show the agent a photo of a shelf or menu, or type out a list of names, and it tells you which option to pick and why. The agent first learns your preferences by asking guided questions the first time you use it, then saves those preferences to editable text files so it remembers them in future sessions. When given an input, the agent identifies which items are beers or wines and ignores everything else, such as spirits or food. It is explicitly instructed not to rely on its own possibly outdated training knowledge, and instead performs a web search on each product to gather details like alcohol content, sweetness, grape variety, or aroma. It then compares each product against your saved taste profile and assigns a preference score from 0 to 100 percent, explaining its top pick, listing alternatives in order, and suggesting food pairings if relevant. Because it is just a text based skill file, it works without any dependencies, subscriptions, or API keys, and can be adapted for use with different AI agents. The README is translated into many languages, suggesting an international audience. This project suits people who already use an AI coding assistant or agent and want a low effort way to get personalized drink recommendations, rather than developers looking for a full application. It is released under the MIT License, so it can be freely used, modified, and shared.