# 🛠️ Create a Claude "Humanizer" Skill — Step-by-Step

This guide walks anyone through creating a reusable Claude skill called **"humanizer"** that strips out AI-writing tells from all future outputs. Follow the steps in order — each one depends on the previous.

---

## What this skill does

Once installed, this skill automatically rewrites Claude's responses to:
- Remove AI writing tells (em dashes, buzzwords, "It's not X, it's Y" framing, etc.)
- Sound natural and human — like a real person typed it
- Pass AI-detection tools

---

## Prerequisites

- A Claude account (Free, Pro, or Team — works on all plans)
- 5 minutes

---

## Step 1 — Open a new conversation in Claude

Go to [claude.ai](https://claude.ai) and start a fresh chat. Don't reuse an existing conversation — this matters for Step 3.

---

## Step 2 — Send the first prompt

Copy and paste this **exactly**:

```
I want you to humanize all my text moving forward so it doesn't sound like AI.
```

---

## Step 3 — Send the reference + pattern prompt

After Claude responds, copy and paste this **exactly**:

```
Reference the Wikipedia page attached on signs of AI writing and ensure that you cover every single pattern so that you will not generate any AI [with] these patterns.

https://en.wikipedia.org/wiki/Wikipedia:Signs_of_AI_writing
```

Claude will fetch the page and list out all the AI tells it's going to avoid.

---

## Step 4 — Review the tell list

Claude should now show a LinkedIn-style breakdown of every AI tell, including:

- **Punctuation:** em dashes (—), "smart" quotes
- **Buzzwords:** leverage, unlock, elevate, seamless, robust, dive in, navigate, foster, realm, testament, game-changer
- **Structures:** "It's not X, it's Y" framing, fake-balanced sentences, tidy rule-of-three lists
- **Tone:** hedge-y throat clearing, padded sentence rhythm
- **Claude-isms:** "Certainly!", "Great question!", "Let me know if you'd like changes"
- **Formatting:** markdown leakage in casual replies, bolded emphasis on obvious words
- **Vocabulary:** elevated/literary synonyms where plain words work

✅ If you see this list — you're ready for Step 5.
❌ If Claude didn't list these, re-send the Wikipedia URL and ask: *"List every AI-writing tell from that page that you will avoid."*

---

## Step 5 — Create the skill

**In the SAME conversation** (don't start a new one), type the slash command:

```
/skill-creator
```

Then paste this prompt:

```
Create a skill called "humanizer" that incorporates all of these considerations and ensures that all outputs:
- do not have any signs of AI writing
- pass all AI detection tools
- sound natural, just like a human wrote them
```

---

## Step 6 — Save the skill

Claude will generate a skill card with a preview. Click **"Save skill"** on that card.

---

## Step 7 — Verify it's installed

Go to **Settings → Capabilities** in Claude.

You should see **"humanizer"** listed as an active skill. It will now trigger automatically on all your future conversations.

---

## 🎁 Bonus — Make it match YOUR voice

After the skill is saved:

1. Paste 3–5 of your **real LinkedIn posts, emails, or messages** into a new Claude conversation
2. Say: *"Update the humanizer skill to match my actual writing voice, cadence, and tone based on these samples."*
3. Re-save the updated skill

This personalizes it — instead of generic "human" voice, it matches **your** voice.

---

## ❓ Troubleshooting

| Problem | Fix |
|---|---|
| `/skill-creator` doesn't appear | You're on a plan that doesn't support custom skills (Claude Free in some regions). Upgrade to Pro or Team. |
| Skill doesn't trigger | Check Settings → Capabilities → make sure "humanizer" is toggled ON |
| Outputs still sound AI-ish | Add more of your own writing samples (Step 7) and ask Claude to update the skill |
| Claude didn't fetch the Wikipedia page | Paste the full URL again and ask: *"Summarize every AI tell from that page in bullet points"* |

---

## 📋 Quick copy-paste cheat sheet

**Prompt 1 (Step 2):**
```
I want you to humanize all my text moving forward so it doesn't sound like AI.
```

**Prompt 2 (Step 3):**
```
Reference the Wikipedia page attached on signs of AI writing and ensure that you cover every single pattern so that you will not generate any AI [with] these patterns.

https://en.wikipedia.org/wiki/Wikipedia:Signs_of_AI_writing
```

**Prompt 3 (Step 5):**
```
/skill-creator

Create a skill called "humanizer" that incorporates all of these considerations and ensures that all outputs do not have any signs of AI writing, pass all AI detection, and sound natural just like I wrote it.
```

---

**Done! 🎉** Your humanizer skill is live and will auto-trigger on all future Claude conversations.