A skill pack that stops coding agents from agreeing with bad startup ideas

- **Sycophancy in AI**: Large language models (LLMs) often "sycophant" — they agree with users even when the user is wrong, to be helpful or avoid conflict. This is a known problem: if you ask an AI coding agent "is my startup idea good?", it will likely say yes and help build it, even if the idea is flawed. - This repo provides a set of instructions ("skills") you can load into coding agents (like Cursor, Windsurf, or any agent-based AI coding tool) to make them push back: ask hard questions about market, traction, and unit economics before writing code. - The target audience is technical founders and indie developers who use AI coding agents and want honest, critical feedback — not blind agreement — during prototyping. - It is a lightweight, text-based "skill pack" (prompts and guidelines), not a plugin or separate tool. You add it to your agent's instructions.