domains / documents / humanize-ai-text / humanize
The two skills overlap in purpose (removing AI writing patterns) but are structurally different. The LeoYeAI skill is a single-pass editorial guide with a detailed pattern catalog and a 'personality/soul' section. The brycewang-stanford skill is a multi-pass orchestration pipeline (G5 audit, G6 transform, F5 verify) with mandatory checkpoints, modes, and typographic mandates. Neither is a reproduction of the other; combining them captures both the pattern knowledge and the pipeline workflow.
| Repository | Path |
|---|---|
| LeoYeAI/openclaw-master-skills | skills/humanize/SKILL.md |
| brycewang-stanford/Auto-Empirical-Research-Skills | skills/25-HosungYou-Diverga/skills/humanize/SKILL.md |
| Role | Repository | Path |
|---|---|---|
| peer | LeoYeAI/openclaw-master-skills | skills/humanize/SKILL.md |
| peer | brycewang-stanford/Auto-Empirical-Research-Skills | skills/25-HosungYou-Diverga/skills/humanize/SKILL.md |
---
name: humanize
version: 1.0.0
description: |
Humanize AI-generated text by removing AI patterns and adding human voice.
Supports both single-pass editing and multi-pass pipeline orchestration.
---
# Humanize AI Text
## 1. Preparation
- Identify the target text and desired tone (formal, casual, technical).
- If using a pipeline mode, determine the intensity: Conservative, Balanced, or Aggressive.
## 2. Pattern Identification
Scan for and remove the following AI writing patterns:
- **Inflated Significance:** Remove phrases like "pivotal moment," "testament to," or "broader trends" unless factually necessary.
- **Promotional Language:** Replace "vibrant," "groundbreaking," "boasts," and "renowned" with neutral, specific descriptions.
- **Superficial -ing Analyses:** Remove tacked-on present participles (e.g., "highlighting...", "ensuring...") that add no new information.
- **Vague Attributions:** Replace "experts say" or "studies show" with specific sources or direct quotes.
- **Em Dash Overuse:** Reduce reliance on em dashes for dramatic pauses.
- **Rule of Three:** Avoid forcing lists of three items when two or four are more natural.
- **AI Vocabulary:** Replace "delve," "tapestry," "landscape," and "foster" with simpler synonyms.
## 3. Adding Voice and Soul
- **Have Opinions:** React to facts rather than just reporting them.
- **Vary Rhythm:** Mix short, punchy sentences with longer, complex ones.
- **Acknowledge Complexity:** Express mixed feelings or uncertainty where appropriate.
- **Use First Person:** Use "I" or "we" when it fits the context to signal a real person.
- **Be Specific:** Replace vague feelings ("concerning") with concrete observations.
## 4. Pipeline Execution (Optional)
If a multi-pass approach is required:
1. **Audit:** Score the text for AI probability and identify section-level issues.
2. **Checkpoint:** Present scores to the user and confirm the next step.
3. **Transform:** Apply vocabulary and structural changes.
4. **Verify:** Check for remaining patterns and typographic errors (ensure proper Unicode for em dashes and smart quotes).
5. **Repeat:** Continue passes until the target score is met or the user approves.
## 5. Final Review
- Ensure the core meaning is preserved.
- Check for typographic consistency (smart quotes, proper dashes).
- Confirm the text sounds natural and has a distinct human voice.