Skip to content
Back to skills

De Ai Ify

ASecurity

Rewrite AI-generated text to sound natural and human-written

  • 52 stars
  • 0 votes
  • 0 copies
  • 0 views
  • Added September 9, 2026
code-qualityapi

Works with

  • cli
  • api

Security analysis

A100/100

Scanned September 9, 2026

npx -y skills add DavidROliverBA/ArchitectKB --skill de-ai-ify --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of De Ai Ify?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for De Ai Ify
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/davidroliverba-de-ai-ify/badge)](https://www.skillsdirectory.com/skills/davidroliverba-de-ai-ify)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: de-ai-ify
context: fork
description: Rewrite AI-generated text to sound natural and human-written
model: opus
---

# /de-ai-ify

Strip AI-sounding patterns from generated text and rewrite to be natural, direct, and human.

## Usage

```
/de-ai-ify <note path>          # Rewrite a specific note
/de-ai-ify --section <heading>  # Rewrite only a specific section
/de-ai-ify --clipboard          # Rewrite text from clipboard context
```

## Instructions

1. **Read the target note** or receive the text to rewrite

2. **Identify AI patterns** — flag any of the following:

   **Filler phrases to remove:**
   - "Let's dive into...", "Let's explore..."
   - "I'd be happy to..."
   - "It's worth noting that..."
   - "It's important to understand..."
   - "In today's rapidly evolving..."
   - "This is a great question"
   - "As we can see..."
   - "Moving forward..."

   **Hedging to reduce:**
   - "It could potentially..." → state directly or remove
   - "This might suggest..." → "This suggests..." or cut
   - "It's possible that..." → state the thing or qualify once
   - Stacking qualifiers: "perhaps", "maybe", "somewhat", "arguably"

   **Structural patterns to fix:**
   - Excessive bullet points where prose would be clearer
   - Every paragraph opening with a transition word
   - Formulaic "In conclusion..." / "To summarise..." wrap-ups
   - Numbered lists for things that aren't sequential
   - Restating the question before answering it

   **Tone patterns to adjust:**
   - Overly enthusiastic ("fantastic", "incredible", "amazing")
   - Corporate-bland ("leverage", "synergise", "drive value")
   - Sycophantic agreement before every response

3. **Rewrite** the flagged sections:
   - Keep the author's intent and meaning
   - Use active voice, concrete language
   - Prefer shorter sentences
   - Let the content speak — cut throat-clearing
   - Preserve any technical accuracy and wiki-links
   - Maintain frontmatter unchanged

4. **Show the diff** — present before/after for each changed section so the user can review

5. **Apply changes** only after user confirms

## Guidelines

- Do NOT change frontmatter, wiki-links, or Dataview queries
- Do NOT remove content — only rephrase
- Do NOT add new content or expand the text
- Preserve the author's voice where it comes through
- When in doubt, cut rather than rephrase
- UK English throughout

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.