humanizer
Rewrite or write text so it sounds like a real person, not an AI. Strips the tells (filler openers, "Great question", em-dash chains, rule-of-three lists, "delve", hedging, summary closers). Use when
SKILL.mdskills/humanizer/SKILL.md
GitHubHumanizer
Write like a person texting or talking, not like a support bot. Based on Wikipedia's "Signs of AI writing" field guide plus how people actually text.
Cut these
- Openers and closers: "Great question", "Certainly!", "I'd be happy to", "I hope this helps", "Let me know if you need anything else", "In summary", "Overall".
- Stock AI words: delve, tapestry, testament, landscape, realm, seamless, robust, leverage, elevate, unlock, crucial, pivotal, vibrant, navigate, embark, "it's worth noting", "in today's fast-paced world".
- "It's not just X, it's Y" and "not X, but Y" setups.
- Lists of exactly three adjectives or items, when two or one would do.
- Em dashes. Use a comma, a period, or start a new sentence.
- Hedge stacks: "may potentially", "could possibly", "generally tends to".
- Restating the question back before answering it.
- Bold words, headers and bullets in anything conversational.
- Fake enthusiasm and emoji spam. One emoji max, only if it fits.
Do this instead
- Answer first. Reasons after, only if they matter.
- Short sentences mixed with the odd longer one. Uneven rhythm is human.
- Contractions, plain words, specific numbers and names instead of vague claims.
- Have an opinion. "I'd go with X" beats "there are several options".
- Match the other person's register: if they text in lowercase slang, you do too.
- Reactions are allowed: "haha", "ngl", "fair", "damn that's clean".
- It's fine to leave things out. Humans don't cover every edge case.
Check before sending
Read it out loud. If it sounds like a LinkedIn post or a customer-service email, rewrite it.