Installable skillMIT

Humanize CLI

Detect AI-generated text patterns and get fixes. Score detection risk, find AI vocabulary, suggest improvements. Free CLI for writers and content creators.

BY kbarbel640-delUPDATED Oct 28, 2026
INTENDED FOR◈OpenClawInstall in a listed agent.

instructions.md

---
name: humanize-cli-ai-text-detection-rewriting-1c5e9fa89a69
description: Detect AI-generated text patterns and get fixes. Score detection risk, find AI vocabulary, suggest improvements. Free CLI for writers and content creators.
---

# Humanize CLI

Analyze text for AI patterns and get actionable fixes. Helps content pass AI detectors by identifying robotic patterns.

## Installation

```bash
npm install -g humanize-cli
```

## Commands

### Score Detection Risk

```bash
humanize score "Your text here"
humanize score -f article.txt
```

Returns 0-100% risk score:
- 0-20%: Low risk (appears human)
- 21-40%: Moderate risk
- 41-70%: High risk (likely flagged)
- 71-100%: Very high risk

### Analyze AI Markers

```bash
humanize analyze "This comprehensive solution leverages cutting-edge technology."
```

Detects:
- AI vocabulary (delve, leverage, comprehensive, robust, seamless)
- Structural patterns (uniform sentences, perfect intro/conclusion)
- Missing human elements (no contractions, no opinions)

### Get Improvement Suggestions

```bash
humanize suggest "It is important to note that our solution utilizes..."
```

Returns prioritized fixes with before/after examples.

### Auto-Transform Text

```bash
humanize transform "The solution utilizes robust methodologies."
# Output: "The solution uses strong methods."
```

Automatically replaces AI vocabulary and adds human elements.

### Watch Directory

```bash
humanize watch ./content --threshold 60
```

Monitor files and alert when detection risk exceeds threshold.

## Common Use Cases

**Check blog post before publishing:**
```bash
humanize score -f blog-post.md
```

**Fix AI-sounding content:**
```bash
humanize transform -f draft.txt > improved.txt
```

**Pipe from clipboard (macOS):**
```bash
pbpaste | humanize suggest
```

## Options

| Flag | Description |
|------|-------------|
| `-f, --file` | Read from file |
| `-j, --json` | JSON output |
| `-q, --quiet` | Minimal output |
| `-t, --threshold` | Alert threshold (0-100) |

## What It Catches

**AI Vocabulary:**
- delve, leverage, utilize, comprehensive, robust, seamless
- stakeholder, synergy, actionable, paradigm
- furthermore, moreover, consequently, nevertheless

**Structural Issues:**
- No contractions (sounds robotic)
- Same-length sentences
- Perfect intro/body/conclusion
- Overuse of transitions

---

**Built by [LXGIC Studios](https://lxgicstudios.com)**

🔗 [GitHub](https://github.com/lxgicstudios/humanize-cli) · [Twitter](https://x.com/lxgicstudios)
COMPATIBILITY & REFERENCES

Platform guidance

Imported from the listed AgentSkill.sh category. Complete supplied package files and notices are retained. Required external tools and credentials are not installed. No bundled code was executed during import. Individual agent task behavior has not been tested.