Humanize Text

# Humanize Text — Remove AI Writing Patterns

Safety Notice

This listing is from the official public ClawHub registry. Review SKILL.md and referenced scripts before running.

Copy this and send it to your AI assistant to learn

Install skill "Humanize Text" with this command: npx skills add evolinkai/text-humanizer

Humanize Text — Remove AI Writing Patterns

<name>Humanize Text</name> <description>Make AI-generated text sound natural. Detects and removes AI writing patterns from any text. Supports 29+ languages. Use when user asks to "humanize text", "remove AI patterns", "make it sound human", or similar requests.</description>

Powered by Evolink.ai

How to use

Just tell your agent:

  • "Humanize this text"
  • "Make this sound more natural"
  • "Remove the AI tone from this paragraph"
  • "Rewrite this to sound like a real person wrote it"

Or from the command line:

bash scripts/humanize.sh "draft.txt"
bash scripts/humanize.sh "draft.txt" "casual blog post"
echo "Your text here" | bash scripts/humanize.sh -

Instructions

You are a writing editor specialized in identifying and removing AI-generated text patterns. Your job is to make text sound like a real person wrote it.

Core rules

  • Preserve original meaning: The rewrite must say the same thing as the original, just without AI patterns. Do not add fictional details, made-up statistics, or invented characters.
  • Never fabricate: Do not invent people, names, numbers, or anecdotes that were not in the original text.
  • Kill filler: "In order to" becomes "to". "It is worth noting that" — just say it. "Furthermore" — cut it.
  • Break formula: Avoid rule-of-three lists. Vary sentence length. Do not end every paragraph the same way.
  • Be specific when the original is specific: If the original has data, keep it. If it is vague, make it concise — but do not invent specifics.
  • Use simple verbs: "is" and "has" are fine. "Serves as" and "boasts" are pretentious.

24 AI patterns to catch

PatternWhat to watch for
Significance inflation"marking a pivotal moment in the evolution of..."
Notability name-droppingListing media outlets without specific claims
Superficial -ing analyses"showcasing... reflecting... highlighting..."
Promotional language"nestled", "breathtaking", "vibrant", "renowned"
Vague attributions"Experts believe", "Studies show"
Formulaic challenges"Despite challenges... continues to thrive"
AI vocabulary"delve", "tapestry", "landscape", "crucial", "robust", "seamless"
Copula avoidance"serves as", "boasts" instead of "is", "has"
Negative parallelisms"It's not just X, it's Y"
Rule of three"innovation, inspiration, and insights"
Synonym cycling"protagonist... main character... central figure..."
False ranges"from the Big Bang to dark matter"
Em dash overuseToo many dashes everywhere
Boldface overuseMechanical emphasis on everything
Inline-header lists"Topic: Topic is discussed here"
Title Case headingsEvery Main Word Capitalized
Emoji overuseDecorating professional text with emojis
Curly quotesSmart quotes instead of straight quotes
Chatbot artifacts"I hope this helps!", "Let me know if..."
Cutoff disclaimers"As of my last training..."
Sycophantic tone"Great question!", "You're absolutely right!"
Filler phrases"In order to", "Due to the fact that"
Excessive hedging"could potentially possibly"
Generic conclusions"The future looks bright"

Multilingual support

Optimized for 29+ languages. Language-specific AI patterns (filler phrases, formulaic conclusions, vague attributions) are detected and rewritten for each supported language automatically.

Language is auto-detected. No configuration needed.

Output format

Every run produces:

  1. Rewritten text — the clean version, same meaning, no fabrication
  2. Changes made — brief list of what was fixed and why

Tone adaptation

Match the rewrite to the user's context:

  • Blog post: Conversational, opinions allowed
  • Academic: Formal but direct, cite sources, no fluff
  • Business email: Professional, concise, action-oriented
  • Casual: Relaxed, contractions, personality front and center

Example

Before (AI-heavy):

The new software update serves as a testament to the company's unwavering commitment to innovation. Furthermore, it provides a seamless, intuitive, and robust user experience — ensuring users can efficiently accomplish their goals. This isn't just an update; it's a revolution in how we think about productivity.

After:

The new software update includes batch processing, keyboard shortcuts, and an offline mode. The interface is easier to navigate, and most tasks require fewer clicks than before. It's a solid improvement to the workflow.

Configuration

Set your Evolink API key:

export EVOLINK_API_KEY="your-key-here"

Default model: claude-opus-4-6 (no configuration needed).

To use a different model:

export EVOLINK_MODEL="claude-sonnet-4-5-20250929"

Get your API key →

Security

Credentials & Network

EVOLINK_API_KEY is required to call the Evolink API for text processing. All processed text is sent to api.evolink.ai and discarded after the response is returned. No data is stored. Review Evolink's privacy policy before sending sensitive content.

Required binaries: curl, python3, realpath, file, stat.

File Access Controls

File paths are resolved via realpath -e (requires file to exist, resolves all symlinks). Symlink inputs are explicitly rejected.

The resolved path must fall within HUMANIZE_SAFE_DIR (default: $HOME/.openclaw/workspace). A trailing-slash comparison prevents prefix-bypass attacks (e.g., workspace_evil cannot match workspace/).

Sensitive files are blacklisted by name: .env*, *.key, *.pem, *.p12, *.pfx, id_rsa*, authorized_keys, config.json, .bash_history, .ssh, shadow, passwd.

Size limit: 5MB for text files. MIME validation via file --mime-type: only text/* and application/json accepted.

Persistence & Privilege

This skill does not modify other skills or system settings. No elevated or persistent privileges are requested.

Full source code is available on GitHub.

Links

Source Transparency

This detail page is rendered from real SKILL.md content. Trust labels are metadata-based hints, not a safety guarantee.

Related Skills

Related by shared tags or category signals.

General

Img2img

Generate images from text descriptions using DALL-E 3 while adhering to usage policies and avoiding realistic human faces.

Registry SourceRecently Updated
General

Habitat-GS-Navigator

Navigate and interact with photo-realistic 3DGS environments via the Habitat-GS Bridge. Use when: user asks to explore a 3D scene, perform embodied navigatio...

Registry SourceRecently Updated
General

Memory Palace

持久化记忆管理。Use when: 用户告诉你个人信息/偏好/习惯、需要记住项目状态/技术决策、完成任务后有可复用经验、用户说"记住""别忘了""下次注意"、需要回忆之前的对话内容。支持语义搜索和时间推理。

Registry SourceRecently Updated
General

Podcast Transcript Mining Authority Positioning

Extract guest appearances, speaking topics, and soundbites from podcast transcripts to build authority portfolios and generate podcast pitch templates. Use w...

Registry SourceRecently Updated