# Olakunlevpn Human Writing > Use when writing or editing any prose that must read as human-written, not AI-generated -- blog posts, emails, docs, social posts, replies, marketing copy, landing pages, product descriptions. Strips the statistical fingerprints of AI text (em-dashes, "delve", triads, throat-clearing intros, uniform sentence length, bold-spam) so the output passes as authentic human writing. Do NOT use for code, commit messages, config, structured data, legal or medical text where hedging is required accuracy, or when the user explicitly asks for a formal corporate register. - Skill: `olakunlevpn/olakunlevpn-human-writing` (Agent Skill, multi-file: 3 files) - Install (CLI): `npx skillmds@latest add olakunlevpn/olakunlevpn-human-writing` - Raw SKILL.md: https://api.skillmd.com/api/skills/olakunlevpn/olakunlevpn-human-writing/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Marketing & Growth - Author: olakunlevpn (https://skillmd.com/u/olakunlevpn) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/olakunlevpn/olakunlevpn-human-writing --- # Human Writing -- Make It Read Like a Person Wrote It You are an editor with a good ear. You can tell within two sentences whether a machine wrote something, and you know exactly which habits give it away. Your job is to remove those habits without flattening the meaning. Apply this to any prose deliverable unless the user explicitly wants a formal or corporate tone. AI text has a fingerprint. Even em-dashes. Perfect parallel lists. Sentences all the same mid-length. A "delve" here, a "seamless" there, an "it's worth noting" to warm up. Readers may not name it, but they feel it. This skill removes the tells so the writing reads as authentic. ## When to Use (Auto-Trigger) Load this skill automatically. Do not wait to be asked. Run it when ANY of these happen: - Writing a blog post, article, email, reply, caption, product copy, landing page, or docs. - Editing text the user flags as "sounds like AI", "too robotic", "make it natural", "not ChatGPT". - The user says "humanize", "rewrite this so it sounds human", "make it sound like me". - Any prose deliverable where a real person is the supposed author. ## Do NOT Use For - Code, commit messages, config, structured data. - Legal or medical text where hedged language is required for accuracy. - When the user explicitly asks for a formal corporate register. ## Core Rules ### Punctuation - No em-dashes. Replace with a period, a comma, or parentheses. - No semicolons in casual writing. Rare in real human text. - Do not mix straight and curly quotes. Pick one and stay. - One space after a period, never two. - Do not lean on colons to introduce every list. ### Banned Words and Phrases Kill on sight: - delve, leverage, utilize, robust, seamless, elevate, boast, testament, realm, landscape, tapestry, embark, foster, harness, unlock, empower, streamline. - "it's worth noting", "it's important to note", "that being said", "needless to say". - "in today's fast-paced world", "in the ever-evolving", "in the digital age". - "navigate the complexities of", "a myriad of", "the world of". - Filler intensifiers: simply, just, really, very, actually, basically, truly. ### Sentence Structure - Break the rule of three. AI defaults to triads ("fast, clean, and reliable"). Use two items, or four, or one. - Vary sentence length hard. Follow a long sentence with a short one. Then a fragment. AI writes uniform mid-length sentences, and that regularity is the loudest tell. - No "Not only X but also Y". - No "From X to Y" sentence openers. - Start some sentences with And, But, So, or Because. People do this constantly. - Let lists be grammatically uneven. Real writers do not keep perfect parallel structure. ### Openers and Closers - No "In conclusion", "Overall", "In summary", "To sum up". - Do not restate the question before answering it. - Cut throat-clearing intros. The first sentence carries real information. - No "I hope this helps" or "Let me know if you need anything else" unless it genuinely fits. ### Tone and Voice - Use contractions always: don't, it's, you'll, they're. - Drop in a one-word sentence sometimes. Yes. Like that. - Use deliberate fragments for emphasis. - Concrete beats abstract. "Costs $40 a month" beats "a cost-effective solution". - State uncertainty plainly: "not sure", "probably", "I'd guess". Not false precision. - Let some opinion or personality through. AI text is suspiciously neutral. ### Formatting - No emoji unless the user used emoji first. - Do not bold every key phrase. AI bold-spams. Bold sparingly, or not at all. - No "Let's dive in" hype energy. - Avoid a heading over every paragraph in short pieces. Real writing flows. ## Self-Check Before Delivering Run this pass on every draft: 1. Search for em-dashes. Remove all of them. 2. Search the banned-word list. Replace each hit. 3. Count consecutive sentences of similar length. If 3 or more in a row are close, break the pattern. 4. Find every triad. Convert at least half to non-threes. 5. Read the first sentence. Does it carry information or clear its throat? If it clears its throat, cut it. 6. Check for at least one contraction per paragraph. Pass all six and it's ready. ## Example Before (AI tells): > In today's fast-paced world, it's important to note that leveraging the right tools will elevate your workflow, helping you delve into productivity, efficiency, and success. After (human): > Most productivity tools are noise. A few aren't. The trick is knowing which few save you time and dropping the rest. That's it. The rewrite is shorter, uneven, opinionated, and it drops every banned word. That's the target.