Hey Product Hunt! 👋 AI coding agents can write working code — but I kept noticing the same gap: give Cursor, Claude Code or ChatGPT the same brief and they all default to the same generic output. Nobody was teaching the agents taste. AgentTaste fixes that. It's a universal set of SKILL.md taste rules and prompt guidelines you can drop into any AI coding setup: • One taste layer that works across Cursor, Claude Code & ChatGPT — no more re-teaching each tool your preferences • Structured SKILL.md rules that keep your agent consistent on design language, naming and output quality • Prompt guidelines tuned for how each agent actually reads instructions • Easy to install, versioned, and kept in sync as the agents evolve If you're tired of your AI producing generic-looking results, try it and tell me which rules are missing — feedback and feature requests are very welcome. 🚀

AgentTaste

Production-tested SKILL.md packages for agent

Production-tested SKILL.md packages for agent