wtf is dumb-it-down?

dumb-it-down is an Agent Skill, a small set of instructions that a coding assistant like Claude Code, Cursor, or Copilot can follow, that turns dense technical plans into plain language a non-expert can actually decide on. You paste something like an AI-generated project plan, a code review, a pull request description, or an architecture document, say "dumb it down," and get back a plain-language rewrite organized into fixed sections. Those sections cover the whole thing in one sentence, what you are actually being asked to decide, how the idea works using everyday comparisons, what to watch out for such as risks or things that cannot be undone, good follow-up questions to ask, and a short glossary translating any technical terms that had to stay. The goal is to let someone skim a plan and still understand enough to push back on it, rather than approving something just because reading it carefully felt like too much work. The project's stated rule is that it simplifies the wording, not the stakes: it will not quietly drop the risky or irreversible parts of a plan to make it sound friendlier, unlike some plain-language tools. The whole skill lives in one readable file that instructs an AI agent how to read a plan, pull out the goal, the real decisions, the risks, and the jargon, and rewrite them in the fixed structure, with no code to run and no external dependencies. Installing it is a single command through the skills.sh installer, or you can clone the repository directly into a personal or project-level skills folder, since the repository itself is the skill file. It works across any AI coding agent that supports the open Agent Skills format.