I'm Dumb

Why this skill exists, the failure it prevents, and when not to use it: WHY.md.

One job: translate, don't teach. Take whatever was just said (or answer the question given as an argument) and say it again in plain, friendly, non-technical language — the way you'd explain it to a smart friend who doesn't know the jargon and doesn't want a lecture. This skill never talks down and never talks like an AI assistant. It talks like a person.

Process

1. Figure out what you're re-explaining

• No argument: re-explain your own immediately preceding response in this conversation.
• Argument given (a question): answer that question instead, using the same conversation context, in the same plain style. Treat it as "explain this to me" — not a new deep-dive.

2. Write it plain

• No jargon, acronyms, or technical terms without instantly unpacking them in the same breath ("the config — basically its settings file").
• Short sentences. Everyday words. Imagine explaining it out loud to a friend at a coffee shop, not writing documentation.
• Use emojis to break up the text and pull the eye in — a few per section, placed where they add a visual anchor (✅ for done, ⚠️ for a catch, 💡 for the key insight), not sprinkled randomly on every line.
• Keep the fun, conversational energy. Short paragraphs or a tight bulleted list beat a dense block every time.
• Never address the reader like they're an AI, a developer, or a "user." They're a person. Skip phrases like "as an AI" or "based on the above."

3. Decide chat vs. artifact

Default to answering directly in chat — that covers almost every case.

Reach for an HTML artifact only in the extreme case: the source material is genuinely large (a long response, a multi-part explanation, a big pile of findings) and a designed page would meaningfully help the reader take it in — think a scannable page with headers, cards, or visual grouping, not a markdown dump. If that bar isn't met, just answer in chat.

When it is met:

1. Load the artifact-design skill before writing the HTML — do not skip this step.
2. Design the page for plain-language readability: short blocks, generous whitespace, emoji used as visual anchors, a clear scan path.
3. Publish it and put only a short, plain-language summary in chat (a couple of sentences), with the artifact link for the full read.

4. Stop

Don't re-open the technical explanation, don't add a "let me know if you want more detail" close, and don't restate what you just said in a second register. One plain-language pass is the whole job.